Skip to main content
Internet Archive's 25th Anniversary Logo

Full text of "Procedures in experimental physics"

See other formats










Experimental Physics 




Experimental Physics 

John Strong, Ph.D. 

Special Research Associate 
Harvard University 

in collaboration with 
H. Victor Neher, Ph.D. 

Assistant Professor of Physics 
California Institute of Technology 

Albert E. Whitford, Ph.D. 

Assistant Professor of Astronomy 
Washburn Observatory, University of Wisconsin 

C. Hawley Cartwright, Ph.D. 

Instructor in Physics 
Massachusetts Institute of Technology 

Roger Hayward 

Architect, Pasadena, California 

Illustrated by Roger Hayward 

New York 

$30. f 


Copyright, 1938, by 

70 Fifth Avenue, New York 





First Printing October 1938 

Second Printing July 1939 

Third Printing September 1941 

Fourth Printing March 1942 

Fifth Printing March 1943 

Sixth Printing June 1943 

Seventh Printing September 1943 

Eighth Printing April 1944 

Ninth Printing July 1944 

Tenth Printing April 1945 

Eleventh Printing November 1945 

Twelfth Printing September 1946 

Thirteenth Printing June 1947 

Fourteenth Printing August 1948 

Fifteenth Printing April 1949 

Sixteenth Printing January 1951 

Seventeenth Printing August 1952 

Eighteenth Printing October 1953 




T is the purpose of this book to describe important pro- 
cedures in experimental physics. Subjects of special 
interest and value to the authors in their own investigations 
have been selected for treatment. Many of the procedures 
and results of research appear here in print for the first time. 

The ideal way to learn the procedures of experimental 
physics is by direct contact with them in the laboratory. 
Realizing this, we have endeavored to bridge the gap between 
laboratory demonstrations and experience on the one hand, 
and exposition on the other, by the liberal use of figures. 

I am indebted to Mr. D. 0. Hendrix for most of the pro- 
cedures presented in Chapter II, and to Dr. R. M. Langer 
^ for the treatment of the unsteady flow of heat presented in 

^ Chapter XII. 

I have drawn freely from many books and scientific 
periodicals. It is hoped that there are no lapses in my effort 
to acknowledge adequately the source of this material. 

The assistance of my wife, of Mrs. Elizabeth H. Hayward, 
and of Mr. James T. Barkelew in the preparation of the 
manuscript is gratefully acknowledged, as is also the help- 
fulness and courtesy of the Prentice-Hall editors. 

J. S. 


Digitized by the Internet Archive 
in 2013 



I. Fundamental Operations in Laboratory 

Glass Blowing 1 

Some physical properties of glass. Cutting tubes and 
bottles. Cleaning. Preheating. The rotation of the 
work. Bending tubes. Shrinking. Annealing. Pulling 
a point. Closing a tube. "Cutting" a tube in the fire. 
Preparations for making joints. Making a joint. Ring 
seals. Blowing bulbs. Constrictions. Correction of 
errors. Platinum seals. Tungsten-glass seals. Copper- 
to-glass seals. Kovar and Fernico. Porcelain-Pyrex seals. 

II. Laboratory Optical Work 29 

Introduction. General procedure. Theory of grinding 
and polishing. Methods of polishing. Procedure for 
optical surfaces of 3 to 6 inches in diameter and larger. 
Cutting and roughing out the work. Biscuit cutter. Glass 
saws. Modified Draper machine. Support of the work. 
Grinding the curve in the work. Fine grinding. Pitch 
for tools. Polishing. Figuring. Cutting zones and transi- 
tion zones. Interpretation of the action of polishing and 
figuring tools. Figuring tools for zones. Manner of 
figuring various zonal defects, and of making aspheric 
surfaces of revolution. Astigmatism. Optical testing. 
Newton's fringes. Haidinger's fringes. Eyepiece tests. 
Foucault knife-edge test. Zonal knife-edge testing. The 
Ronchi test. Hartmann's test. Lining up a system of 
mirrors. Two methods of generating optical surfaces. 
Working optical surfaces on the hand-lever machine. 
Relationship between two optical surfaces. Blocking. 
Quartz and calcite. Optical working of crystals. Polish- 
ing of metals. The Schmidt camera. 

III. Technique of High Vacuum 93 

The law of ideal gases. The mean free path. Viscosity 
and heat conductivity. Pumping speeds. Conductance of 
vacuum pumping lines. Evacuation. Roughing pumps. 
Outgassing of glass and metals. Vapor pressure of waxes. 
Getters. Static and kinetic vacuum systems. Diffusion 
pumps. The use of oils as diffusion pump liquids. Oil 
diffusion pumps. Mercury traps. Virtual leaks. Oil 
traps. Construction of kinetic vacuum systems. Joints. 
Seals. Electrodes. Valves. Mechanical motion. Leaks. 
Vacuum gauges. The McLeod gauge. The ionization 
gauge. The Pirani gauge. The Langmuir gauge. The 
Knudsen gauge. 



XIV. Notes on the Construction and Design of 

Instruments and Apparatus .... 569 

The cutting of metals. The lathe. Soft soldering. Hard 
soldering. Spot welding. Instrument design. Vibration- 
less supports. 

X"V. Molding and Casting 593 

The lost-wax method. Patterns for sand casting. Sand 
casting. Cuttlebone casting. 


Fundamental Operations in Laboratory 
Glass Blowing 

r T , HE fundamental operations in glass blowing for labor a- 
■*■ tory use are cutting, rotating, bending, blowing, and 
welding. By various combinations of these operations, 
apparatus is constructed from glass tubing and glass cane. 
It is the purpose of this chapter to describe how these opera- 
tions are executed. Hard glass, such as Pyrex, is now used 
extensively for making laboratory apparatus. It is more 
difficult to manipulate than soft glass because it has a higher 
working temperature and thus congeals quickly when it is 
removed from the flame. However, less difficulty is ex- 
perienced in annealing hard glass because of its low thermal 
expansion and high strength. Since this more than out- 
weighs the greater skill required for manipulation, we will 
be concerned chiefly with hard glass in this chapter. 1 
v An arrangement of a glass-blowing workbench is shown in 
Fig. 1. Cross-fires are shown for heating the glass to soft- 
ness, a method that may be termed American, since German 
glass blowers ordinarily use a single-blast burner. Com- 
pared with the blast burner, cross-fires heat the glass more 
rapidly and uniformly. Either method may be used for 
most of the operations. However, some of them require 
a pointed flame, which is more easily obtained with a blast 

Here, where we treat of the American method, the hand 
torch, mounted as shown by the dotted lines in Fig. 1, is 
used to obtain the pointed flame. Natural or artificial gas 

1 Glass may be obtained from the Corning Glass Company, Corning, New 



[Chap. I 

is used for fuel in the burners. Compressed air is used for 
working soft glass ; but in order to obtain the higher tempera- 
ture required to work hard glass, oxygen or a mixture of 
oxygen and air must be used. In an ordinary blast burner, 
however, acetylene can be used as fuel with compressed air. 

hand torch may be hung 
on cross -fires for 


tubing support 
(wood with 
metal weights) 

ra.cK for assorted 


tape — 

» iwooaen 
if supports 
— i_ for work 

to cool 

table high enough 
for operator to rest 
his elbows on 

Fig. 1. 

Accessory equipment includes a collection of corks of 
various sizes, some fitted with closed glass tubes to serve as 
handles for rotating the work, and others with open tubes 
for blowing. Pieces of rubber hose of various sizes fitted 
with closed glass tubes, to close up the ends of small tubes, 
are also included. A swivel L and mouthpiece device with 
a connecting rubber hose, shown in Fig. 1, is convenient for 
blowing rotated work that is large or otherwise awkward to 

Chap. I] 


bring to the mouth. Forceps and molding tools used for 
spinning glass are shown in Fig. 2. A file for cutting small 
tubes and a hot-wire device for cutting larger tubes are 
shown in Fig. 3. To sharpen the corners of the file, the 
narrow sides are ground on an emery wheel. When the file 

nichcl %6'thich 


S inches or more 
Fig. 2. 



tight fit 

set with 


V» cement 


extra wire 
for adjusting 

to sUe of -tube to be cot 



[Chap. I 

requires tempering, it is heated until it becomes a dull red 
and plunged into cold mercury. 

Pyrex tubes of various sizes, capillaries, and cane are kept 
in stock for constructing apparatus. There should also be 
a supply of other glasses, such as soda glass, lead glass, and 






2 3 

Non metals 

+ ■»- 

+ -+ 

^gsS^tT^ Fused Quartz 

lOO 200 30O 400 500 600 700 800 

temperature in degrees C. 

Fig. 4. 


Nonex. These should be well labeled and kept apart 
from the main stock. 

Some physical properties of glass. The thermal expan- 
sion of various glasses and metals is shown in Fig. 4 
and Table I. Other characteristic temperatures of glass 
and quartz are given in Table II. The variation of viscosity 

Thermal Expansion Coefficients 


Expansion Coefficient 
(X 10 7 ) 

Lime glass (G8) 


Lead glass (G5) 


Nonex (G702P) 


Porcelain (20° to 290° C.) 


■ Pi 

Quartz glass (16° to 1000° C.) . . . . 








80 to 100 

61 to 65 

Fernico and Kovar: 

25° to 450° C 


25° to 500° C 






Characteristic Temperatures for Glass and Quartz 


Strain Point 



Soft glass 

389° C. 

425° C. 



750° to 1100° C. 


1756° to 1800° 



[Chap. I 

with temperature for a typical glass is shown in Fig. 5. 
The viscosities corresponding to important characteristic 
temperatures — annealing temperature, working tempera- 
ture, and melting temperature — are indicated on the curve 
in Fig. 5. The significance of the first two temperatures is 


>"*" state 


viscous state 

(brittle) ihere a strain 

devitrifying . 
"" range 


iwill disappear 
'in about four 


i 1 

+ + 

limit at which 
| viscosity is 
I great enough 
-f to prevent 
I crystallization 

►- > • temperature > > - 
Fig. 5. 

that internal strain is relieved in about 4 hours when glass 
is heated to the yield point, while only about 4 minutes are 
required at the annealing temperature. At the yield point 
the viscosity is about 10 13 poise. At the annealing tempera- 
ture it is about 10 12 poise. In the working range of tempera- 
ture the viscosity varies between the limits 10 5 and 10 10 
poise, with the optimum working viscosity about 10 86 poise. 

Chap. I] 


Glass is considered molten when the viscosity is less than 
10 2 poise. 

Cutting tubes and bottles. To cut small glass tubes 
(to \ inch in diameter) for the operations of glass blowing, 


file mark 
Fig. 6. 

they are first scratch-marked with the sharp edge of a file, 
care being taken that the scratch, a few millimeters long, is 
accurately perpendicular to the tube. A break is then made 
by a combined bending and pulling force as illustrated in 
Fig. 6. Tubes can be broken at the scratch-mark by means 
of a stroke with the file as 
shown in Fig. 7. This tech- 
nique is suitable when the 
tube is hot or when it is to 
be cut near the end. 

Tubes larger than J inch 
in diameter require a differ- 
ent technique. After being 
scratch-marked with the file, 
they may be cracked by 
applying the tip of a small 
piece of glass cane, made in- 
candescent in the flame, to 
one end of the file mark. 
The crack thus produced 
may or may not completely encircle the glass. If not, it can 
be made to do so by leading it with repeated applications of 
the glowing cane tip, each application being just ahead of 
the end of the crack. 

Fig. 7. 


A tube or bottle of several inches in diameter is cut by 
first filing a narrow scratch-mark around its circumference. 
A piece of stiff paper or cardboard may be used to guide the 
file in making this mark. The wire of the device shown in 
Fig. 3 is adjusted to fit in the mark. The ends of the wire 
must not touch. An electric current is passed through the 
wire, heating it to a red heat for a few seconds, and water is 
applied to the scratch-mark and wire with a pad of wet 
cotton. This procedure will produce a clean crack around 
the circumference. Small irregularities in the crack may be 
removed by grinding on a brass plate with Carborundum 
grits, or after the glass has been softened in the flame they 
may be pulled off with forceps or cut off with shears. 

Cleaning. Good welds cannot be made with contaminated 
glass. Therefore, the first operation after cutting should be 
cleaning. Sometimes washing with water is sufficient, but 
nitric acid may be substituted if necessary. In extreme 
cases, hot chromic acid "cleaning solution" may be required. 
Water used to rinse glass tubing is removed from the out- 
side with a clean cloth and from the inside with a wad of 
cotton pulled through with a string or blown through with 
air. Or, if distilled water is used, the tube may be dried by 
drawing air through it with a water aspirator and by warm- 
ing it gently at the same time. 

Preheating. Glass tubing and especially large glass 
apparatus must be preheated carefully before they can be 
safely exposed to the local intense heat of the cross-fires or 
hand torch. By one procedure for preheating, the work is 
first exposed to the relatively cool flame of a Meker burner 
with the air shut off. As the glass temperature rises, more 
and more air is admitted to the Meker burner, giving a hotter 
and hotter flame, until finally, when the work is thoroughly 
heated, it is safe to expose it to the intense ^heat of the 
cross-fires or blast burner. By an alternative procedure the 
work is exposed to the heat of the cross-fires for a fraction of 
a second, after which it is quickly withdrawn to allow tem- 
peratures to equalize, and then after a few seconds another 

Chap. I] 


section of the work is exposed. This operation is repeated in 
such a way that the temperature of the work as a whole is 
uniformly elevated. The exposure to the flame is increased 
and the interval outside the flame decreased as the heating 
progresses, until the work is brought to a temperature at 
which it distills enough sodium vapor to make the flame 
yellow. This sodium test usually indicates a temperature at 
which it is safe to begin the operations of shrinking, blowing, 
molding, and so forth. Some things, such as tubes, require 
preheating only in the zone around the region to be worked. 

The rotation of the work. Rotation of work is a funda- 
mental operation. It should be executed uniformly and 
with good coordination of the two hands. Glass properly 
rotated in the flame becomes uniformly soft, and the effect 
of gravity on it is symmetrical. 

The lower surfaces of hot glass cool more rapidly than the 
upper surfaces. For this reason it is also important to con- 
tinue uniform rotation even after the work is removed from 
the flame. 

Fig. 8. 

The beginner will have difficulty manipulating the work 
in the flame, particularly after the glass connecting the two 
parts on either side of the flame becomes soft, when he may 
"tie up" the work. To avoid this, he is advised to practice 
rotation with a model consisting of two glass tubes connected 
with fairly heavy cloth. He should be able to rotate these 
in the manner shown in Fig. 8, so that the cloth does not 



[Chap. I 

wrinkle or twist and is under neither compression nor ten- 
sion. He is then ready to begin operations with the flame. 
The work is manipulated by the thumbs and forefingers 
so that, despite differences in diameter, the sections of the 
work on either side of the soft zone in the flame are rotated 
in synchronism, the motion consisting of a series of angular 
displacements of about 45°. The left hand always handles 
the heavier section of the glass, while the right manipu- 
lates the section beyond the soft zone. The right hand 
has the more delicate though lighter task, since it must ro- 
tate its section in phase and without undesired stretching 
or compression relative to the main section of the work. 
The hands are held as shown in Fig. 8 to facilitate the appli- 
cation of the right end of the work to the lips for blowing. 
Bending tubes. A tube to be bent is heated in the cross- 
fires with continued rotation until it is quite soft along a 

length equal to several di- 
ameters. It is then removed 
from the flame and bent to 
the desired angle with the 
apex down as shown in Fig. 9. 
As large tubes are difficult to 
heat uniformly, imperfections 
often occur. They are also 
present in small tubes, par- 
ticularly in small, thin-walled 
tubes, that have been bent 
to a sharp angle. Imperfec- 
tions are worked out in every 
case by local heating with a 
pointed flame. When one 
portion of the tubing wall is 
heated until it is soft, the 
general form of the bend is maintained by the portion on 
the opposite side of the axis of the tube. If the outside 
tends to flatten as shown in Fig. 9(b), it is corrected by 
blowing while the glass is soft. If the inside surface folds 


Fig. 9. 



hand torch 
hung up to 
I eave both 
hands free. 

Fig. 10. 

as shown at (c), it is locally heated with a sharp pointed 

flame and worked by alter- n^tai mandrel 

nating shrinking with blow- V . _^ wr ^S^ t6 T th 

ing until it is uniform. These 

corrections are followed by a 

general heating to anneal the 

whole bend. 

A glass coil is made on a 

mandrel. The mandrel is 

usually either a steel or brass 

tube covered with asbestos 

paper. The paper is applied 

wet, the ends being lapped 

and cemented with sodium 

silicate. After the paper is 

dry, this lap joint is sand- 
papered. One or more coats of stove polish or some other 

form of carbon will prevent the glass from adhering to the 

asbestos. Notches in the end 
of the tube secure the coil to 
the mandrel. The procedure 
is illustrated by Fig. 10. 

Shrinking. Since softened 
glass is a liquid, its surface 
tension tends to deform it in 
such a way that the total sur- 
face is decreased. Shrinking 
at elevated temperatures is 
restrained by the viscosity of 
the glass, and this restraint is 
greater at the lower limit of 
the working range. Shrinking 
may yield both desirable and 
undesirable changes in the 
work and it is controlled by 

the use of spinning tools and by blowing into the work. 

Fig. 11 shows the use of forceps to counteract the undesirable 


The cut end of a tube 




when heated gets thicker, 

*= — = 


and by spinning' it ovt 


one may form a reinforced 
rim for a stopper. 

Fig. 11. 


tendency of the end of a tube to decrease its diameter by 
shrinking, while the desirable effect of increased wall thick- 
ness is achieved. 

Annealing. The annealing of complicated and elaborate 
work is one of the most difficult operations in glass blowing. 
It is also an important one, since, if the work is not properly 
annealed, it may break in cooling or, what is worse, fail 
after it is put into operation. The purpose of annealing is 
to bring the glass from the working temperature to room 
temperature with the introduction of a minimum amount of 
strain. Annealing is properly executed when all parts of 
the work are maintained at a uniform temperature while the 
glass is gradually cooled. Large, complicated work should 
be annealed in a suitably regulated oven. Small work in 
which the wall thicknesses are uniform can be successfully 
annealed either with a Meker burner or in the cross-fires. 

When the manipulations have been completed, the work is 
heated until it is above the annealing temperature. The 
temperature is then gradually lowered by applying the pro- 
cedures of preheating in reverse order. It is important that 
the temperatures be kept uniform during the cooling by 
special extra applications of heat on those parts which tend 
to cool more rapidly, either because they are thinner or be- 
cause they are subject to greater heat losses by radiation and 
convection. When the temperature is judged to be well 
below the strain point, the work may be set aside for final 
cooling in a place free from drafts. 

Pulling a point. "Pulling a point" is a technical term used 
by glass blowers indicating that a tube is heated in the flame 
and drawn out as illustrated in Fig. 12 to give a "point," 
which is usually some 6 inches in length. The point may 
have several functions. It may serve as a handle for rota- 
tion or, with the tip removed, as a mouthpiece through which 
to blow; or it may afford a means of closing the work. Also, 
pulling a point is a preliminary to several other operations. 

We will assume, for the purpose of our discussion here, that 
a section of tubing is required with points on both ends as an 



element of some apparatus under construction, and further- 
more that this is to be obtained from a longer stock tube. 
First, a point is pulled on the end of the stock tube. If it 
is long, the stock tube may be supported on the left by a 


~"" — — — — ^ ^^~ 

rt^.r, cane 

(gloss rod) 

(f)( J # hZ^ removing excess tip 

(g) (Pr ~k 
7 ^\ 

thickening the end 


^ end blown out to hemisphere 
Fig. 12. 


V-block as shown at (a) in Fig. 12. After preheating the 
tube by the second procedure outlined above, it is softened 
a few diameters back from the end. Then the glass is 
gathered together at the tip with forceps; the work is re- 
moved from the flame, and with continued rotation the soft 
glass is drawn out as shown at (b). The capillary section is 
fused in the middle as shown at (c), or, if the point is to serve 
as a mouthpiece, it may be cut and fire-glazed by momentary 
exposure to a flame. 

The tubing is then heated until it is soft at a suitable 
distance back of the first point, and the desired section is 
drawn off, forming at the same time the second point. 

It is important to have the walls of the point symmetrical 
about the axis of the tube. Errors may be corrected by 
heating the shoulder of the point until it is soft and manipu- 
lating the glass from the end of the capillary. It is advisable 
to work the glass at a low temperature when making cor- 

Closing a tube. Pulling a point is the first operation in 
closing a tube as shown in Fig. 12(d) to (h). The point is 
removed with a sharp flame as shown at (d) and (e). Excess 
glass at the tip is removed with forceps or with a piece of 
cane (f), and the end is then heated to shrink it (g) ; then it is 
blown to the final hemispherical shape (h). The hand torch 
is usually used for this operation. 

"Cutting" a tube in the fire. The first step in "cutting" 
a tube in the fire is to pull a point. Again the point is re- 
moved as described above, Fig. 12(d) and (e), and excess 
glass removed, Fig. 13(a). The end is then heated (b) 
and blown with a strong puff to yield a thin kidney-shaped 
bulb (c), which is broken off with the file or forceps as shown 
at (d). The edges are now heated to shrink them and 
thicken them to the size of the tubing elsewhere (e). The 
diameter is increased by a spinning process and the use of 
forceps as in (f) or flanging tool as in (g). If forceps are 
used, they are introduced and allowed to expand slowly as 
the glass is rotated in the fire. The end of the tube is then 

Chap. I] 



squared with the carbon plate (h). If a flange is required, 
the end is spun out with the arrowhead spinning tool and 
squared with the carbon plate as shown at (h) and (i). 
Metal spinning tools are wet with beeswax to prevent stick- 
ing to soft glass. 

removing excc&s 

heating end 

blown out 

bulb broken off 

thickening end 

-SweHinq to 

,;■'/) \\\\ uniform oYiameter* 



squaring off 
carbon plate 


Fig. 13. 



[Chap. I 


^_ ML •' 

Fig. 14. 

Preparations for making joints. 

Thorough cleaning of the glass tube 
and careful attention to the pre- 
liminaries of cutting, flanging, 
rrzzgi or drawing and expanding it 
;-v f ac iiit a t; e th e manipulations 
in the flame. A common fault 

in the beginner is that he thinks he can easily correct defi- 
ciencies in these operations 
after the work is in the flame. 
It is significant that good 
glass blowers do not handi- 
cap themselves by careless- 
ness with these preliminaries. 

The elements that are to 
be welded to form a joint 
must have approximately the 
same diameter and wall 
thickness. If a large tube is 
to be joined to a smaller one, 
the large tube is first pre- 
pared as shown in Fig. 14(a) 
by pulling a point on it and 
then cutting off the point in 
the flame where the shoulder 
has the same diameter as the 
small tube. 

A capillary or thick-walled 
tube is prepared as shown at 
(b). It is heated to softness 
and blown until it has the 
proper wall thickness and 
then pulled until it has the 
same diameter as the tube 
to which it is to be sealed. 

A bulb or cylinder to which 
a small tube is to be joined is Fig. 15. 

edges thickened 
carbon rod 







Chap. I] 






Fig. 16. 

first preheated. A soft flame is then directed on the place 

selected for the joint until it is soft, and a slight bulge is 

blown as shown in Fig. 15(a). 

This bulge is strongly heated at 

its apex with the tip of a sharp 

flame as at (b). Then, after re- 
moving it from the flame, a small 

thin-walled bulb is blown as at 

(c). This is then broken off with 

the forceps or file. The edges of 

the hole thus made are softened 

with the flame, flanged with the 

carbon taper, and squared with 

the carbon plate as shown in 

Fig. 15(d) to (h). 
A straight tube is prepared for 

making T's by opening the side 

as described above. When sev- 
eral T's are to be made, a holder for the straight-through 

tube as shown in Fig. 16 is con- 
venient. Y's are made by first 
bending a tube to an acute angle. 
This is then opened at the apex as 
shown in Fig. 17. 

Making a joint. The elements 
are heated with rotation in a flame 
whose diameter is approximately 
the same as the diameter of the 
tubes. They are arranged facing 
each other, as shown in Fig. 18, 
with the axis of the joint perpen- 
dicular to that of the flame. When 
the tubes are thoroughly soft at 
their ends, they are removed from 
the flame and touched together at 

right angles as shown at (b). This contact is used as a 

hinge to steady the hands while the tubes are brought 

Fig. 17. 



[Chap. I 

into exact register and pushed together (c). With con- 
tinuous rotation, the joint is held in the flame until it 
shrinks to a uniform outside diameter (d). It is then with- 
drawn from the flame and blown out until it has a uniform 

Fig. 19. 

wall thickness (e), and stretched at once to a uniform out- 
side diameter (f). Since it is necessary to blow a joint, 
obviously all openings except the one applied to the lips must 
be temporarily closed. 

Large tubes that are to be joined must have flanges. 
When it is necessary to make a joint on apparatus which 
cannot be rotated, the squared ends of the elements of the 
joint are accurately fitted together and heated, a section of 
the circumference at a time. The welding of the flanges is 
effected with the heat of the hand torch and pressure applied 

Chap. I] 



with the forceps working around the circumference as shown 
in Fig. 19. After this the joint is locally heated, a small 
section of the circumference at a time, until it is soft, and the 
softened area is worked by alternate shrinking and blowing 
until the wall is smooth. Then the whole circumference is 
uniformly heated for final blowing, alignment, and annealing. 
Ring seals. When a tube is inserted in a bulb or a larger 
tube, a ring seal joins the tubing wall to the edge of the aper- 
ture in the bulb or large tube. First, the glass around the 

blown ring 

point opened 
for blowing 

plug to support 
inner part 






to bold inner tubes in line 

Fig. 20. 

aperture is accurately molded with the carbon taper until 
it is slightly larger than the outside diameter of the small 
tube to be inserted. The small tube is prepared for the seal 
by heating a narrow zone around its circumference with a 
pointed flame and swelling it as shown in Fig. 20(a). This 



[Chap. I 

is accomplished by blowing and simultaneously applying a 
longitudinal compression. The small tube is then inserted 
and held exactly concentric with the larger tube by an im- 
provised support, such as a roll of asbestos paper, as illus- 
trated in Fig. 20(b). The place to be sealed is exposed to a 
pointed flame with continued rotation until the glass at the 
ring is soft. Then the weld is made by pushing the swelling 
of the smaller tube against the constricted opening of the 
larger tube. The work is removed from the flame, blown, 
and aligned, while at the same time the small tube is given a 
slight pull. Fig. 20 shows the construction of a water 
aspirator which requires two ring seals. A tapered wooden 
dowel which just slips into the first tube centers the second 
while it is being sealed. Ring seals require careful annealing. 
Another procedure for ring seals, particularly suited for 
inserting a small tube through the side of a larger tube, is 

illustrated by Fig. 21. The inner section of the insert is 
flanged and molded to conform to the inside wall of the large 
tube and is supported in contact with it as shown at (a). 
The area of the outside wall of the large tube opposite the 
place where the section makes contact on the inside is then 
heated until the two tubes are sealed together. A bulge is 

Chap. I] 






blown and opened with a sharp flame at the center of this 
seal as shown at (b). The opening is molded and a small 
side tube is joined to the edges of it to form a continuation of 
the inner section as shown at (c) and (d). 

Blowing bulbs. Difficulty may be experienced in making 
large bulbs (of 2 inches in diameter or larger), for which it is 
necessary to heat heavy masses of soft glass to a uniform 
temperature in the flame. Also, the work must be skillfully 
managed outside the flame to make 
the effects of air cooling symmetri- 
cal. Because of these difficulties 
it is advisable to use commercial 
balloon flasks for bulbs rather than 
to make them from tubing. Small 
bulbs, less than 1 inch in diameter, 
are not so difficult to make. 

The first operation in making 
bulbs is to heat the end of a glass 
tube until the glass collects as 
shown in Fig. 22(a). As glass col- 
lects, it is alternately blown out 
and shrunk to distribute it uni- 
formly until enough has collected 
for the final bulb. The collected 
glass is then heated to a uniform 
temperature and removed from the 

After the work has been rotated a few seconds about a 
horizontal axis, it is expanded by blowing through an ap- 
propriate mouthpiece. The blowing is gentle at first and 
stronger as the glass stiffens. The work is continuously 
rotated. However, if one portion of the surface tends to 
expand more rapidly than the other portions, it is turned 
down and cooled to restrain its expansion, since the under 
side of the work cools most rapidly. 

To blow a bulb in the middle of a tube, the operation of 
collecting glass, as described above, is carried out in zones 

Fig. 22. 


until several adjacent ones are obtained as shown in 
Fig. 23(a) to (c). Then, by blowing and shrinking, these are 

. combined in a single 

GEZ ~^^ C ZZL J U uniform zone (d). This 

* ' is well heated, removed 

from the flame, allowed 
to cool a moment, and 
blown to the desired 
form (e). 

Constrictions. Two 
types of constrictions 
may be required. One, 
useful for preventing 
kinetic overflow of mer- 
cury in a manometer 
tube when the pressure 
suddenly changes, has 
a constricted inner wall 
but uniform outside di- 
Fig 23. ameter. The second 

type, useful as a "seal- 
on " for a vacuum system, has a uniform wall thickness. To 
make either type, the glass tubing to be constricted is heated 
and worked until the glass walls thicken. This operation is 
essentially the same as the preliminary operation for blowing 
a bulb in the middle of a tube as shown in Fig. 23(a). After 
the walls have been thickened, the glass is removed from the 
flame, and the tube is rotated and pulled instead of blown as 
for a bulb. To get a constriction of the first type, the tube 
is pulled until the outside diameter of the tube is uniform, 
while to get a "seal-off" constriction the tube is pulled until 
the wall thickness is uniform. 

Correction of errors. Owing to errors of manipulation, 
the walls of glass apparatus frequently are not uniform. 
This lack of uniformity not only detracts from the workman- 
like appearance of the finished apparatus but also increases 
the difficulty of annealing, since the thick and thin portions 


tend to cool at different rates, a circumstance which causes 
strain in the glass. 

Excessive glass can be drawn off from a region in the walls 
of an apparatus by using a piece of cane as illustrated in 
Fig. 13(a). After the required amount of glass is drawn off, 
the region is worked by blowing and shrinking until the wall 
thickness becomes uniform. Also, if the wall of a region is 
too thin, glass can be added from a piece of cane and worked 
out smooth by blowing and shrinking. Holes may inad- 
vertently appear in the work. They are closed by drawing 
their edges together with a piece of cane. 

Platinum seals. Formerly, the only satisfactory method of 
making a metal-to-glass seal was by the use of platinum 
and soft glass. Such seals are rarely used now because of the 
high price of platinum. Also, hard glass, which seals directly 
to tungsten, is now used extensively for making laboratory 
apparatus. However, a plat- _ >> 

mum tube may be required to platinum j be<xd Keated 
introduce pure hydrogen by w>r ^^t JL * 

diffusion into a glass apparatus, f \ - - ■ 
For this and other special pur- 
poses, platinum-soft glass seals 
are required. 

Fig. 24 shows a platinum 
electrode in a soft-glass tube. 
In making this platinum seal, a 
small bead of soft glass (either 
lead or soda glass) is first fused 

, ? . . finished seal 

to the platinum wire. The Fi 24 

bead and wire are heated to 

about 1000° C. to obtain a good glass-to-metal bond. Then 

the bead is sealed into the wall of the tube as shown in the 


Tungsten-glass seals. Tungsten wires may be sealed 
through Pyrex if their diameters are less than 0.060 inch. 
Larger tungsten wires, to diameters of twice as much, are 
first sealed in a sleeve of Nonex glass, which in turn is sealed 



[Chap. I 

into the wall of glass apparatus. This latter operation is 
necessary, especially if the seal is to be exposed to the heat 
of a baking-out oven. Nonex glass has a lower softening 
temperature than Pyrex, and between the strain point and 
room temperature the total thermal expansion of Nonex is 
almost equal to the expansion of tungsten for the same tem- 
perature interval. 

A tungsten wire is prepared for sealing through glass by 
heating it to a white heat in the gas flame. If this is not 

done, bubbles appear at the 

tungsten up to 

Pyrex or 
nonex tubingf 

yA drop of 

r n 

nickel or 
fused to 

copper braid -^ J 

fused to advance drop' 

(b) finished seal 
Fig. 25. 

surface of the seal. The sur- 
face of the tungsten is cleaned 
for sealing by heating and 
touching it with a piece of 
potassium or sodium nitrite. 
The tungsten is then washed, 
and a short sleeve of Pyrex 
(or Nonex, depending on the 
,____ size of the wire) tubing is 

-^^^pJ^^^ fused to it as shown in Fig. 
J-C--, - Z ^y ^% 25(a). The intense heating 

required to shrink the glass 
should be started at one end 
of the sleeve, so that the 
shrinking progresses from that end. This avoids trapping 
air bubbles between the metal and the glass. The interface 
between glass and tungsten is red, because oxide on the 
surface of the tungsten dissolves in the glass and dyes it. 
After the sealing operation between glass and metal is 
finished, the sleeve is welded into the apparatus as shown 
at (b). In making metal-to-glass seals, it is important to 
cool the glass slowly to avoid excessive strain. 

Tungsten wire is frequently fibrous, having longitudinal 
channels which may leak if it is sealed into a vacuum 
apparatus. To avoid such a possibility, the tip of the tung- 
sten should always be closed by fusing nickel or advance 
wire over it. The nickel or advance tip also affords a place 

Chap. I] 



for attaching copper wires. Copper can be fused to these 
tips, whereas it cannot be easily welded to tungsten directly. 
Discharge tube electrodes are made from coiled aluminum 
wire of about -^ inch in diameter and a tungsten-Pyrex seal 
as shown in Fig. 26. The aluminum-wire projection of the 
coil is fused to make the connection to the tungsten wire. 
The tungsten wire with a nickel enlargement to secure it in 


nickel tip 
welded on to 
prevent wire 

copper foil 

Pyrex or 



eJuminum-wire eleetrode 


er wire 

tube is drawn down hi 
and. .second electrode is 
shaken into place for 
second seed 

Fig. 26. 

position is pushed into the fused aluminum. The projection 
is wrapped with copper foil to preserve its form. After the 
aluminum has solidified, the copper foil is removed. A 
glass sleeve, shaped as illustrated, is then sealed to the 
tungsten. This sleeve fits the aluminum projection and 
affords additional support for it. 

Copper-to-glass seals. It is possible to seal copper to Py- 
rex or soft glass by the technique developed by W. G. House- 
keeper. 2 The copper has a much larger coefficient of thermal 
expansion than either type of glass — it is the arrangement 

2 Housekeeper, VY. G., Elect. Engineering. 42, 954 0923) 



[Chap. I 

i Pyrex tubes 

copper t-ube 
reamed -to thin 

ring of Pyrex 
fused into 
mouth of 

tubes fused j 
to copper^/ 



^score with 
fiko.nd break 

hole drilled in 
copper dish 

Pyrex tube 



copper wireV 


^ Pyrex must not 
run over edge 
of copper 

( Pyrex tube fused 
around copper 

=?l~) wire silver 
\M soldered 

through disk section 

Fig. 27. Housekeeper glass-to-metal seals. 

hammered flat 

of the seal which prevents the glass from breaking. When 
the copper is thin, it is deformed to absorb differences be- 
tween its expansion and that of the glass, a circumstance 
made possible by its high duc- 
tility and low yield point. The 
construction details of various 
seals developed by House- 
keeper are shown in Figs. 27 
and 28. For the constructions 
shown in Fig. 27 it is impor- 
tant to prevent the glass from 
passing over the rim of the 

The seal shown in Fig. 28 
is made with a copper wire, 
which is hammered out to have 
thin sharp edges. Care is 
necessary in heating the Pyrex 
to avoid melting the copper. 

— side 
— **— Section 

i nserted 
in hot Pyrex tube 

end of tube flattened 


finished seal 
Fig. 28. Housekeeper seal. 

Chap. I] 




Kovar and Fernico. 3 The rate of expansion of glasses 
increases near their softening temperatures, as Fig. 4 shows. 
On the other hand, the thermal expansion for most metals is 
nearly linear. However, the expansions of two new alloys, 
Kovar and Fernico, closely duplicate the expansion of some 
of the commercial glasses. 4 These alloys yield metal-to- 
glass seals which are un- 
strained under all annealing 
conditions, and they may be 
sealed to appropriate glasses 
without any of the special 
procedure required for 
Housekeeper seals. Large 
seals of 4 inches in diameter 
and | inch in wall thickness 
have been made between Kovar and 705 A J glass. Such 
seals as the ones shown in Fig. 29 have made modern all- 
metal radio tubes possible. 

^spot weld 

Fig. 29. 

3 The fundamental study of expansion properties of Fe-Ni-Co alloys, on 
which this kind of metal-to-glass seal is based, was made in the Westinghouse 
Research Laboratories by Howard Scott, Technical Publication 318, American 
Institute of Mining and Metallurgical Engineers (1930). These alloys are 
manufactured under U. S. Patent 1,942,260, held by the Westinghouse 
Electric and Manufacturing Company. Further information is contained 
in Scott, Howard, Frank. Inst., J., 220, 733 (1935); Burger, E. E., Gen. El. 
Rev., 87, 93 (1934); and Hull, A. W., and Burger, E. E., Physics, 5, 384 
(1934). The Westinghouse product, called Kovar, is obtainable from the 
Stupakoff Laboratories, 6627 Hamilton Avenue, Pittsburgh, Pennsylvania. 
Fernico is obtainable from the General Electric Company, Schenectady, 
New York. 

4 According to A. W. Hull, "Fernico is capable of existing at room tempera- 
ture in either the gamma, face-centered phase, or in the alpha, body-centered 
phase. When annealed from 900° or more, it has the face-centered structure 
and the characteristic low expansion, and is stable in this condition at any 
temperature above —40° C. Exposure to liquid air temperature or me- 
chanical strain will transform it into the alpha phase, which has a different 
expansion and is to be avoided." 

According to Mr. Scott, "To obtain the desired low and reversible expansion 
characteristic of Kovar and Fernico, their composition is so adjusted that 
transformation from the gamma to alpha phase occurs between — 80 and 
— 180°C. Seals, however, cannot be cooled safely below — 40°C. because of 
the progressively increasing expansion between metal and glass on cooling 
below room temperature. Special compositions can be made which make 
possible cooling to somewhat lower temperatures." 



[Chap. I 

These new alloys may be soft-soldered, copper-brazed, 
and spot-welded. It is not recommended that they be 

silver-soldered, however, as 
this tends to make them 
brittle. They oxidize much 
less than iron and therefore 
do not oxidize seriously at 
elevated temperatures. 
Nevertheless, care should be 
taken to avoid extended 
overheating during sealing. 
An important property of 
the alloys is that they are 
not attacked by mercury. 

Porcelain-Pyrex seals. 
Porcelain, particularly the 
grade known as Insulite, 5 
may be sealed directly to 
Pyrex in small diameters 
(less than \ inch), or it may be sealed to Pyrex in large 
diameters with an intermediary glass ring of Nonex as 
shown in Fig. 30. 

Fig. 30. 

5 Insulite is obtainable from Stupakoff Laboratories, 6627 Hamilton Avenue 
Pittsburgh, Pennsylvania. 


Laboratory Optical Work 

Introduction. In this chapter we will describe the tech- 
nique of making the optical surfaces required for mirrors, 
prisms, interferometers, lenses, and so forth. The optical 
surfaces on these instruments are characterized by being 
much more accurate than ordinary machined and ground 
surfaces. In fact, optical tests sensitive to a few millionths 
of an inch are necessary to show their lack of true perfection. 
Since our concern here is with high-precision work, in which 
errors are usually less than a wave length of light, we do not 
include methods used for plate glass, cheap lenses, and other 
commercial work in which the tolerance is greater. 

The technique described here is intended primarily to 
guide the research worker who finds it desirable or necessary 
to prepare his own optical surfaces. . 

In any case the task set before the worker is that of 
generating an extremely accurate polished surface. Ac- 
cordingly, we will first set down the general technique in- 
volved without detailed reference to what is being made. 
Later we will treat of special procedures. 

General procedure. The glass or other material on which 
the optical surface is to be prepared is first roughly formed to 
the desired shape. For example, in the case of a lens, the 
first step will consist of cutting out a disk of glass. A prism 
will be first sawed or ground to rough dimensions from a 
larger block. The proposed surface itself is then generated 
more precisely by periods of grinding with suitable laps. 
The surface is ground first with coarse grits of Carborundum 
to conform approximately to the specifications. Then 
finer and finer abrasives are used until at last the grinding is 
terminated with the finest emery flour. The grinding is 




[Chap. II 

periodically interrupted for testing with a straightedge, 
template, micrometer, or spherometer. After fine grinding, 
the surface is polished with a pitch lap and rouge. Finally 
it is brought as near to perfection as the specifications re- 
quire by "figuring," that is, by local retouching with polish- 
ing tools. The figuring is guided by delicate optical tests. 

Theory of grinding and polishing. Optical grinding and 
polishing are alike in that both require the use of a material 
which is harder than the glass. This material is used in the 
form of loose grits or fine powder. The two operations are 
unlike in that the grits and powder used for grinding are 
worked over the surface with a hard tool, ordinarily made 
either of glass or cast iron, whereas polishing tools are made 
from a soft base material. A polishing tool for preparing 
precise optical surfaces is usually composed of some combi- 
nation of pitch and wax as the soft base material. Paper 
cloth and wood are often used for polishing tools in cases in 
which no great precision is demanded. 

The grinding process depends upon the characteristic 
conchoidal fracture produced when an excessively high 
pressure is applied to a point in the surface of the glass. 


particle of 

fracture starting 


tension forces 

on / > 

compression I h 
forces *%^^* 

Fig. 1. 

The pressure exerted on the surface by a single particle of 
abrasive or grit, as it is rolled about between the tool and the 
work, builds up stress beyond the strength of the glass, 
resulting in the removal of a chip. This is illustrated by 
Fig. 1. Carborundum and emery grits are ordinarily used. 
The efficiency of the process depends primarily on the sharp- 
ness of the grits. Carborundum grits break down faster 

Chap. II] 



than emery, although they are harder than emery. Frac- 
tured Carborundum grits have sharp edges and consequently 
they grind fast. Ellison 1 says that Carborundum grinds 
about six times as fast as emery. Carborundum is used for 
the first coarser grades of grits, and emery for the last finer 
grades. Natural emery (corundum) cuts about twice as 
fast as synthetic emery. The corundum produces a smoother 
surface than either Carborundum or synthetic emery and is, 
accordingly, best for the final grinding. 

The hardness of various abrasives is indicated in Moh's 
extended hardness scale. (See Table I.) 

Hardness Scales 

Moh's Scale of Hardness 

Extended Moh's Scale 







Orthoclase or periclase . . 

Vitreous pure silica 













Fused zirconia 


Fused alumina 


Silicon carbide 


Boron carbide 




Ridgway, R, R., Ballard, A. H., and Bailey, B. L., "Hardness Values of 
Electrochemical Products," a paper presented before the Electrochemical 
Society, May, 1933. 

From a practical point of view, we may consider that the 
polishing operation is a planing process. 2 The grains of 

1 Ingalls, Albert G., editor, Amateur Telescope Making, page 74. New York: 
Scientific American Publishing Company, 1935. 

2 For a more comprehensive treatment of the theory of polishing from a 
different point of view, see the following: 

Lord Rayleigh, Proc. Opt. Convention, No. 1, page 73 (1905); and Scientific 
Papers, Vol. IV, page 542. Cambridge: The University Press, 1903. 

French, J. W., "The Working of Optical Parts," Dictionary of Applied 
Science, Vol. IV, page 326. London: The Macmillan Company, 1923. 

Finch, G. L, "The Beilby Layer," Science Progress, 31, 609 (1937). 



[Chap. II 

abrasive appear to fix themselves automatically in the soft 
material of the tool, usually pitch, so that their crystal 
surfaces are parallel to the direction of motion of the tool 
and parallel to the plane of its surface. Thus a complex 
scraper is formed. As this moves over the glass, the height 
of each abrasive particle is automatically adjusted in the 
soft backing so that it produces a fine smooth cut. The 
removed glass is washed away by the liquid lubricant, 
usually water. The planing action starts on the peaks of the 
" hills" that result from the fine grinding and produces a full 

polish there at the first stroke. 
fi ^L£ r ° und su / facc J^ ^ Continued operation of the 

polishing tool removes addi- 
tional glass, so that the hills 
become plateaus and are fi- 
nally planed down to the level 
of the deepest valleys. The 
character of the surface on 
any particular plateau is not 
improved by continued pol- 
ishing — it is to be regarded as 
fully polished from the first 
stroke. This is illustrated in 
Fig. 2. After the whole sur- 
face becomes uniformly pol- 
ished, further working with the polishing tool removes 
additional glass. In constructing an experimental aspheric 
camera lens, as much as thirty thousandths of an inch of 
glass has been removed by polishing. 

Methods of polishing. Glass can be successfully polished 
with almost any fine abrasive, provided a suitable soft and 
yielding backing is used. For some types of work — for 
example, for edging mirrors where irregularities in the sur- 
face do not matter — glass is polished with a wood tool 
charged with Carborundum or emery. Glass may be pol- 
ished with rouge, either the red oxide or the magnetic black 
oxide, and also with charcoal or oxide of tin. However, for 

about .0002 inches 

polished surface *. \ £_ 

Fig. 2. 


ordinary optical work rouge is the most satisfactory polish- 
ing material. Surfaces of glass, quartz, speculum metal, 
calcite, and flaorite are best polished with rouge on a wax or 
pitch tool. The action of various polishing agents depends 
on the type of backing, whether cloth, paper, or pitch is used, 
on the hardness of the material being polished, and on the 
method of lubrication. Some agents which are indifferent 
polishers when used with a wax or pitch tool and lubricated 
with water are quite effective when used dry on a paper lap. 
For paper polishing, oxide of tin (putty powder) is commonly 
used. Chromium oxide (Cr 2 3 ) is recommended for polish- 
ing certain metals such as stainless steel which are " attacked" 
by rouge. 

The material for the polishing tool may be a soft metal — 
copper, lead, or aluminum. Tools made of these metals 
are sometimes used for polishing thin specimens of minerals 
which are to be examined with the microscope. Levigated 
alumina is usually employed as abrasive for work of this 

The polishing tools used in precision optical work are made 
of pitch or pitch and wax compounds, in contrast with cloth- 
or paper-faced tools used on some commercial products. 
Glass is polished with surprising rapidity on a cloth polisher, 
but it exhibits a peculiar grainy " lemon-peel" surface. This 
method of polishing is generally used in the manufacture of 
plate glass. Paper polishers in general produce a somewhat 
better surface than cloth but are seldom used except for the 
manufacture of inexpensive lenses, such as for cheap hand 
magnifiers and so forth. All polishing tools of a fibrous 
nature produce a "lemon-peel" surface. 

Procedure for optical surfaces of 3 to 6 inches in diameter 
and larger. The technique which will form the nucleus of 
our first treatment is particularly suited to the making of 
surfaces of 3 to 6 inches in diameter or larger worked in 
glass or quartz. The procedures involved are fundamental 
and apply equally to mirrors, lenses, or prisms. The 
method treated here is used by D. 0. Hendrix, a practicing 



optician associated with Mount Wilson Observatory. 3 This 

procedure is different in some respects from that described 

, , . , in the classic book on ama- 

steel-wheel glass cutter . . , . . . 

teur telescope making by 

Ingalls, Porter, and Ellison. 4 

For example, in their book 

they recommend using the 

tool underneath the work, 

while here we treat primarily 

of the method using the tool 

on top of the work. 

Cutting and roughing out 

the work. The work, whether 

it is a mirror, a lens, or a 

prism, can often be cut to 

rough shape from stock plate 

glass with an ordinary wheel 

cutter, the most common 

form of glass cutter, which is 

used for cutting all kinds of 

polished glass in all ordinary 

thicknesses. The cutter is 

drawn across the glass surface 

once with sufficient pressure 

definitely to mark the glass. 

It should not be run back and 

forth along the same line. 

After the glass is "marked," 

it is broken by bending it 

away from the cut, as shown, 

for example, in Fig. 3. The 

parallel-jawed pliers, also illustrated in Fig. 3, are useful for 

making narrow cuts. The break may also be started by 

lightly tapping the glass on the back side opposite the mark 




Fig. 3. 

3 1 am indebted to Mr. D. O. Hendrix for the procedures presented here. 
4 Ingalls, Albert G., editor, Amateur Telescope Making. New York: Scien 
tific American Publishing Company, 1935. 

Chap. II] 



with the small knob provided on the handle of the wheel 

The procedure for cutting 
thick plate glass is to lubricate 
the wheel cutter with turpen- 
tine or kerosene before the cut 
is made. After the glass is 
marked, the break is started 
with a blunt chisel. The chisel 
is held firmly against the back 
of the glass at a point directly 
opposite the mark and tapped 
sharply with a small hammer. 
The edge of the chisel should 
be parallel to the mark. It is 
well to have the glass sup- 
ported, cutter-mark down, on 
a cloth or padded surface. 
When the break has started, 
it is led along the cut with the chisel. 

A hammer 
and chisel 
>s used to 
cad" the 

made with 
a glass 


iron disk fed 
with Carbo- 
rundum and 

pan to 
catch the 

Fig. 5. 

(See Fig. 4.) 

Very thin glass is best cut 
with a diamond point espe- 
cially mounted and sharpened 
for this purpose. 5 

If a disk is desired, the glass 
is first cut square, and the 
corners are then cut to give a 
polygonal piece approximating 
the desired shape. The rough 
edges may be removed by hold- 
ing the glass against a rotating 
flat disk of cast iron fed with 
a mixture of Carborundum and 
water. (See Fig. 5.) Also, 

5 Diamond glass cutters may be obtained from the Standard Diamond Tool 
Corporation, 64 West 48th Street, New York City. This company also 
sharpens diamond glass cutters. 





iron tool 

Be careful to protect the 
lathe from the abrasive 
which will spatter from 
the vibrating tool. 

Fig. 6. 

the glass disk may be waxed onto a metal plate mounted 
in the headstock of a lathe. As it is rotated, the edges are 
ground with an iron tool, which is fed with Carborundum 

.feed with Carborundum and water as shown in Fi S- & 
A light springy tool is recom- 
mended. Oilcloth should be 
used to protect the lathe so 
that abrasive does not get into 
its working parts or on its 

Biscuit cutter. A common 
method of cutting small 
disks (up to about 6 inches 
in diameter) from slabs of 
glass is by means of a "biscuit 
cutter." This is simply a 
thin-walled tube of iron or 
brass mounted in a drill press 
as shown in Fig. 7. The ro- 
tating tube is fed against the glass, Carborundum and 
water being applied with a spoon. Fig. 7 also shows a 
novel method of central feeding. To prevent chipping 
when the biscuit cutter goes 
through the glass, it is well 
to wax an auxiliary backing 
plate onto it with beeswax. 
Grade 60 or 90 Carborundum 
should be used except for fine 
cuts or cuts on delicate and 
fine work, in which case grade 
120 Carborundum should be 
used. The cutter will cut more 
rapidly if instead of water a 
mixture of turpentine and cam- 
phor is used with the Carborundum. The proportions 
of the mixture should be 5 grams camphor to J liter tur- 

Fig. 7. 

Chap. II] 



Diagonal mirrors, such as the Newtonian diagonal for a 
small telescope, may be cut out of a larger figured flat mirror 
with the biscuit cutter. The larger mirror is mounted in the 
drill press at an angle of 45° and cemented with beeswax 
on a backing of plate glass. It is cut in the manner shown 
in Fig. 8. Usually the front of the flat is also coated with 
a cover glass, stuck on with 



to prevent 




Fig. 8. 

beeswax, to prevent scratch- 
ing the figured surface with 
the abrasive. A mirror thus 
cut out may develop a slight 
turned-up edge. However, if 
the cut rim, which now has 
a rough ground surface, is 
polished with wood and Car- 
borundum, the figure of the 
mirror will usually become 
flat again. (This is some- 
times called the Twyman 

Glass saws. Strips and slabs of glass are cut from a thick 
piece of glass stock with saws. The simplest and easiest of 
these to set up is the so-called "mud saw," shown in Fig. 9(a). 
It consists of a rotating disk of soft sheet iron fed with a mix- 
ture of Carborundum and water. Sometimes sugar, syrup, 
talc, glycerin, or bentonite (particularly good) is added to 
this mixture to make the Carborundum adhere to the blade 
and to keep the grits from settling out in the reservoir pan. 
The usual construction allows one edge of the saw to dip 
into the "mud," or Carborundum mixture, which is held in a 
pan below the disk. The work to be slabbed is supported on 
a counterbalanced table and is held against the saw with a 
slight pressure. 

A diamond saw forms an efficient slabbing cutter. The 
diamond saw shown at the bottom of Fig. 9 may be made as 
follows: The diamonds are pulverized as shown at (b) and 
charged into the nicks of a circular disk prepared as shown 


gloss being cot 

[Chap. II 

hardened steel ^--^ 

mortar and pestle FTf 

for pulverizing J 


After the diamonds are 
pulverized they are mixed 
with wax. The dust should 
be between 80 and 100 mesh. 
A ten-inch saw will need 
about 4 carats of crushed 
bort (rough diamonds). /.v 

The nicked edge of the 

copger disk is 

filled with 

wax and 



A yi6-inch- 

thick disk 

of soft 

copper is 

supported on arT 

arbor between steel plates. 


The nicks are spaced V^2 W 
to V\e apart. 

glass being cut 

tank for 
water or 

The nicks 
are closed 
by knurling 
with a smooth steel roller. 
( d ) 

The closed nicks 
grip the particles of diamond. 

hardened steel roller 
ll^rolled edge of the disk 

H If the edge does not 
i / l swage out as shown, 
the sides of the disk 
must be faced off to 
give clearance. 

The edge of 
^^ the saw should 
run at about 
looo feet per 

Fig. 9. 


at (c). These nicks are rolled as shown at (d) to hold the 
diamond powder and give the saw clearance. In operation 
the saw blade is lubricated and washed with water or 

Modified Draper machine. Once the prism, lens, mirror, 
or other blank is cut out, the operations involved in grinding 
the curves and polishing and figuring them may be carried 
out either by hand or with a grinding and polishing machine. 
A machine like the one shown in Fig. 10, a so-called modified 
Draper machine, is suitable. 

The tool is moved laterally by the modified Draper 
machine in a thin oval stroke across the face of the work. 
The amplitude of this stroke is controlled by adjustment of 
the throw of the crank. The stroke can be arranged by 
movement of an adjustable guide so that it is either diame- 
tral or chordal in respect to the work. The tool may be 
allowed to rotate freely, or it may be driven by a belt. Also, 
the tool may be loaded to increase its pressure, or it may be 
counterbalanced to decrease its pressure on the work. The 
table on which the work is mounted is power driven to rotate 
about 2 r.p.m. 

Support of the work. It is very important to support 
the work properly, or it will develop astigmatism, the 
anathema of optical work. 

The first requisite is to have the modified Draper ma- 
chine table turned and lapped 0.001 to 0.003 inch concave, 
depending on the size. It is then covered with a layer of 
thin felt and oilcloth as shown at the bottom of Fig. 10. 
This supports the glass uniformly on its flat bottom side and 
effectively prevents flexure during all of the operations. 
When the second surface of a lens is being worked, the plane 
concave glass tool that was used in the fine grinding of the 
first face is used to support the work on the grinding table. 
The tool is first mounted on the table concave side up. 
Then it is covered with felt and the lens is laid on it. 

The work is supported laterally on the table by three edge 
arcs, which should fit neatly to the edge of the blank without 

crank with 

loose brass sleeve 

The weight may be 
placed here to reduce 
the pressure on the 
tooj when desired. 

cotter pins 


This pulley should 
be driven by a ^.-h.p. 
motor through appropriate 
Speed reductions. 

^t> wooden edge arcs 

two layers 
of oilcloth 

fe* felt 


" ^ "~ J ^ cwq yeast-iron turntable 

with top surface 
ground flat 

Fig. 10. Note that the universal joint shown here is used only for rough 
grinding. For fine grinding and polishing, the tool is connected to the crossarm 
by a pivot and socket arrangement. See Figs. 16 to 2C 



exerting any pressure on it except as is necessary to balance 
lateral forces produced by the action of the tool. The work 
is moved around in the edge supports from time to time dur- 
ing grinding and polishing to distribute the effect of these 
forces uniformly around the periphery of the mirror, to 
avoid the introduction of astigmatism. 

The work, if it is a mirror or lens, is prepared by having 
its face and back fine-ground and made parallel with a ro- 
tating cast-iron lap used with loose grits as shown in Fig. 5. 
The edges are then ground round and lightly beveled. 
Finally, the edges are polished with a wood tool and fine 
Carborundum grains. 

Grinding the curve in the work. Full-size grinding tools 
of tough metal such as copper, brass, or soft iron, when 
turned to a definite radius of curvature, will reproduce this 
radius in the glass. The soft metal surface becomes charged 
with abrasive and is not worn appreciably when it is used on a 
brittle material such as glass. On the other hand, cast-iron 
tools change slowly during grinding, and glass tools change 
at approximately the same rate as the work. 

The traditional way of making a 6-inch mirror by hand is 
to use two equal disks of glass, one as the work and the other 
as a grinding tool. The grinding is accomplished as shown 
in Fig. 11 with the work mounted on a firm pedestal, the 
height of which is optional. The optician walks around it 
as he strokes the work with the tool. Pressure is applied to 
the center of the tool with the thumb of the right hand. 
The tool is rotated with the fingers in a counterclockwise 
direction as it is stroked across the right side of the work. 
When a chordal stroke is used, the upper disk becomes 
concave and the lower convex. By this means a certain 
amount of control is given the operator. He may continue 
grinding, increasing the curvature in the surfaces all the time, 
until the desired result is attained. If he wishes to decrease, 
the curvature, he will place the tool below and stroke it 
with the work. Or he may periodically reverse the relative 
positions of the two disks if he wishes to hold the surfaces 



fairly flat or constant in radius of curvature. When it is 
desired to hold the curvature constant, a diametral rather 
than a chordal stroke is used. 

On the modified Draper machine, the grinding of a mirror 
to a definite radius of curvature is effected with a small tool. 

edge, arcs 

handle for the work 


.avy carpet 01 
sponge rubber 

As the operator strokes the 
work he slowly rotates the 
part in his hand as he walks 
slowly around the stand. 

Fig. 11. Many workers will prefer to have the work at a lower level, 
3 to 4 feet, than is shown here. Note : The operator shown in this figure is 

Concave curves are cut in the glass with a J-size tool strok- 
ing the work across its center. A convex curvature is gen- 
erated by a sub-diameter tool stroked across a chord of the 
work. Although a convex curvature will be generated if a 
full-size tool is stroked across the center of the work (di- 
ametral stroke), it becomes convex more rapidly when a 
chordal stroke is used. The rate at which the curvature 
changes is proportional to the amplitude of the diametral 
stroke or the offset of the chordal stroke. 


After the work has been roughed out to the proper radius of 
curvature with 90, or for extreme curves 60, Carborundum, 
the full-size tool is used to true up the surface. The stroke 
used here is a thin oval across the center of the work. The 
amplitude used is about one third to one sixth the diameter 
of the work. The grinding is continued with the full-size 
tool until the tool and work are spherical. This is indicated 
by the quality of the fit between the tool and the work, 
which can be tested with a pencil mark made on the work. 
This procedure may produce scratches. A circular template 
is often made of the required radius and the work is ground 
until it fits this template. Spherometers are also used to 
test the work for sphericity. When the work is spherical, 
the spherometer reading, d, the radius of curvature of the 
work, R, and the radius of the circle containing the spher- 
ometer legs, r, are related as follows : 

, r 2 + d 2 

d = ~2lT' 

The spherical surfaces obtainable by grinding are so good, 
in fact, that opticians who worked before testing methods 
were developed as they are today hesitated to polish the 
grinding pits entirely away, since they formed a convenient 
" landmark" to which to refer the figure. 

To grind deep curves like those required for an//l Schmidt 
camera, one puts a band around the edge of the mirror and 
covers its face with a layer of Carborundum grits. The 
band holds the grits on the mirror. As the work slowly ro- 
tates, a fast rotating sub-diameter cast-iron ring tool is 
reciprocated diametrically, or nearly so, across its surface 
in a thin oval stroke. The amplitude of the stroke is ad- 
justed so that the ring tool comes to the edge of the work at 
the extremes of the stroke. 

Final grinding in all cases should be carried out with a 
glass grinding tool. Glass is used rather than metal in order 
to have the tool change at approximately the same rate as 
the work, thus insuring a more perfect fit at all times between 



[Chap. II 


the tool and the work. The tool may be a glass disk formed 
as a complement of the work; that is, if the work is a convex 
spherical surface of radius R, the tool will be a concave 
sphere of almost exactly the same radius. Or the tool may 
be a plate of glass cemented to a metal backing. It is well 
to cut the grinding tool with one or more decentered grooves 
as shown in Fig. 12 in order to prevent suction, facilitate the 

access of grinding compound 
to all parts of the tool, and 
insure that the tool grinds 
slightly faster than the work. 
These grooves may be cut into 
the glass with the diamond or 
mud saw. 

For large mirrors, glass disks 
or squares can be cemented 
to a convex or concave iron 
backing as is illustrated in 
Fig. 12. 

The radius of curvature of 
the work, R, is determined by 
means of a spherometer or 
more simply by a template 
cut from metal. The latter 

grooves ><32 
to y t & inch 
deep* cut 
saw or 
diamond saw 

glass grinding tool 

glass facets 
cemented to 
curved iron tool 
with hard pitch or 


grinding tool for large work 

Fig. 12. 

can be cut with a sharp steel point (sharpened like a brass 
turning tool) mounted on the end of a board of length R 
and pivoted at the other end on a nail. For flats a good 
straightedge may be used as a template. 

Fine grinding. After the proper radius is attained and 
the work has been trued up with the full-size tool, the optician 
passes successively to grades 150, F, 400, and 600 Carbo- 
rundum. The full-size tool, loaded to a pressure of about 
0.5 lb. /square inch, is used. For a 6-inch mirror about a 
teaspoonful of grits is applied at a time. Each application 
of grits, applied with one or two spoonfuls of water, is 
allowed to grind until the gritty cutting sound, which is 
heard at first, has softened. For a 6-inch mirror, grits are 


repeatedly added until the work has been ground a total 
time of 30 minutes (or 1 hour by hand). After a half -hour 
of grinding with one grade of Carborundum, the optician 
passes on to the next grade, and finally, after the 600 grade 
Carborundum, finishes with two grades of emery, 302J and 
303 J. The work, the table of the machine, and the tool 
should be thoroughly washed after finishing with each grade 
of abrasive. 

Carborundum grits as obtained commercially are well 
graded and do not need to be washed. However, the emeries 
must be washed each time they are used. The washing 
procedure is as follows : Put emery to a depth of 1 inch in a 
quart Mason jar, fill the jar with water, stir, and let settle 
for 10 seconds. Decant the suspended emery off to a second 
clean jar and discard the residue. After 10 seconds in the 
second jar, decant again, and repeat the operation a third 
time. After this, the settling time is increased to a minute 
to yield a residue which we will designate as residue A. 
The liquid over this is decanted into a clean jar, in which it 
is allowed to settle until it is clear, yielding residue B. 
The liquid over B is then put back over residue A, stirred, 
allowed to stand for 1 minute, and then added again to B. 
This is repeated several times to transfer a large fraction of 
the emery from A into B. Residue B, when mixed with an 
equal volume of powdered washed talc, is ready to be used 
for grinding. The talc serves as a lubricant and prevents 
sticking of the tool. The talc must be washed in the same 
manner as the emery was washed to free it of metallic 

The final grinding with the two grades of emery will yield 
a surface which exhibits specular reflection of white light at 
grazing incidence. At a steeper angle the reflected image is 
red. In fact, specular reflection of the red part of the spec- 
trum up to a grazing angle of about 12° may be obtained. 
The maximum grazing angle of specular reflection affords a 
simple test of the quality of the fine-ground surface. A 
clear filament lamp should be used as a light source for this 


test, and when the surface gives a reflection at a grazing 
angle of about 12°, the work is ready to be polished. 

When it is required to have the center of the mirror 
perforated, the necessary hole is usually cut with the "biscuit 
cutter" before the grinding is started. The plug is then 
fastened back in place with plaster of Paris. The plug is 
left in place until the figuring is finished. 

Pitch for tools. Polishing pitch should have the follow- 
ing properties: It should flow slightly at ordinary room 
temperature; it should trim easily with a sharp knife; and, 
further, it should not lose its "temper" by evaporation of 
volatile oils. A compound which conforms to these specifi- 
cations quite well is made up as follows: 

Coal tar (melting point 170° to 180° F.) 2 lbs. 

Pine tar (Mefford Chemical Company) 4 liquid oz. 

Beeswax 1^ oz. 

Venice turpentine not more than about 2 or 3 cc. 

The tar is melted and the other ingredients are added in the 
order listed. 

The function of the turpentine is to adjust the final 
"temper" of the pitch. More or less turpentine is added, 
depending on whether a hard or soft pitch is desired. Before 
adding the turpentine and after each addition, test the pitch 
for temper. The simple method of performing this test is to 
chew a small sample of the pitch after chilling it by pouring 
it out on a cold glass surface. At body temperature, so- 
called "soft" pitch can be chewed, while "hard" pitch cracks 
under the pressure of the teeth. Furthermore, hard-pitch 
tools stored face up will show evidence of flow in the sides 
of the groove in about a week. A soft tool exhibits flow 
after standing a day. Polishing pitch does not attain its 
final hardness on cooling but continues to harden for a day 
or more. This is a sort of "jelling" process, which must be 
taken into account. 

After the correct mixing temper is attained, pitch is 
filtered to remove small sticks or other hard particles. The 
hot pitch is poured through a cheesecloth filter supported on 


an iron ring. Two layers of cheesecloth are adequate to 
hold back harmful impurities. 

Polishing tools of 6 inches in diameter or less are made by 
simply pouring the melted pitch compound over a support 
to a depth of about f inch. After the pitch has cooled, it is 
channeled by cutting it with a hot knife so that the surface is 
divided into a decentered system of square facets of uniform 
size. These facets are later trimmed in the manner shown in 
Fig. 13. Tools having bubbles in the pitch cause no trouble 
unless it happens that the bubbles occur in a definite zone 
on a full-size tool that is to be worked over the mirror or lens 
with a short stroke. In order to avoid a zone of bubbles, the 
pitch is cast by pouring it onto the support at one edge rather 
than at the center. 

There are two methods of accommodating the tools to 
the different working conditions of summer and winter: 
By one, the formula is changed, the pitch being tempered 
with more turpentine for cooler weather; by the other, the 
size of the facets is changed. The facets are made smaller 
in cooler weather. When the formula given above is used, 
the facets should be about 1 inch square for temperatures 
above 75° F. and about J to f inch square for temperatures 
below 75° F. 

If the polisher is to be used on soft or easily scratched 
material, such as speculum metal, it is advisable to use 
harder pitch and to have the facets narrow. The channels 
allow the pitch to flow evenly and also allow the rouge and 
water free access to all parts of the work. For speculum 
metal it is recommended that the facets be -^ to tV inch 
wide and } inch long. 

To construct a polishing tool of relatively short radius, 
the pitch facets are first cast in a suitable mold in the form 
of sticks. (See Fig. 14.) These are then cut into squares 
and fastened to the metal tool as shown in Fig. 15. 

After the tool has been faceted, it is warmed and pressed 
to the work, with soap in a 25 per cent glycerin solution as a 
lubricant on the work to prevent sticking. The pressing 







paper rim 
^/ string 

Hot pitch 
will not stick 
to the dry 
Joe on gummed 
paper tape, 
^glass or metal -tool 

warmed before pourino" 
the hot pitch 
hot Knife blade - % A 6 - 

v ' '■■** w.M**\ strap iron 
16 inches 

paper lining 

go entirely 
the pftch 


3 /4"tol" 




pitch between 


the cuts 
are trimmed 
with a sharp 



polishing tool for glass 

tool for 

sticks, lightly 
tacked in place 

mold for casting strips 
of pitch h 

filling the 

facets % /i* by 1* 

The strips of 
pitch when 
from the 
mold are 
cut into 

Fig. 13. 

Fig. 14. 

Chap. II] 




The pitch 

facets are 
warmed and 
stuck onto the 
warmed tool. 

operation is illustrated in Fig. 16. The tool is gently warmed 
over a hot plate until the pitch is soft. Then it is applied to 

the work, wet with a mixture 
of soap and glycerin, and left 
to cool. This procedure yields 
an intimate contact between 
the tool and the work. Tools 
for flats may be first turned in 
the lathe before they are 
pressed. After pressing the 
pitch tool, it is advisable to 
wash it in cold water and also 
wash and dry the work to re- 
move the soap and glycerin. 

Polishing. Polishing is car- 
ried out on the table of the 
modified Draper machine in 
the same manner as grinding, 
except that the polishing tool 
is usually allowed free rotation. Rouge and water is added to 
the work from time to time near the edge of the tool with an 

typical polishing tool for 
work of short radius 

Fig. 15. 

The back 
of the pol- 
ishing tool 
is slightly 
and then the 
pitch is 
until it 
to the 
firm pres 
sure of 
the thumb- 

The work 
is mois- 
with a 
few drops 
of soap / 
and glyc J 
erin to 

\ ^he polish- 

W^^ing too! 

is pressed 

firmly on 

/the work 

rAfor a. 
\ two and 
then left 
until cool. 

piece of 

the work 

Fig. 16. 

eye dropper. The rouge should be washed. The washing pro- 
cedure is the same as that described for washing emery or talc, 
except that the settling time is longer — up to one-half hour. 


Hard facets in the tool may cause sharp zones to appear in 
the work during polishing. To avoid the effects of such 
surface inhomogeneities in the pitch and resultant irregu- 
larities in the cutting action of the tool, the work is "broken 
up"; that is, the tool is frequently (and irregularly) given a 
spin. The facets in the tool should form a decentered sys- 
tem. A diametral stroke is employed for polishing. A 
feature of the polishing machine which also contributes to 
breaking up the work is the incommensurable coupling ob- 
tained by the belt which connects the rotation of the work 
with the phase of the stroke. The stroke is varied from 
time to time from a long stroke of one fourth the diameter 
of the tool to a short stroke. 

If the surface of a pitch polishing tool becomes so heavily 
charged with rouge that it appears hard and glassy, the 
polishing speed will be considerably reduced, and further- 
more "sleeks" are liable to appear. Sleeking, or the appear- 
ance of groovelike marks on the polished surface, is probably 
caused by the formation of ball-shaped aggregates of rouge, 
wax, and perhaps glass, which plow out shallow channels 
in the surface. Beeswax-coated tools are particularly 
bothersome in this respect. One method of avoiding sleeks 
is to allow the tool to run nearly dry before each application 
of fresh rouge. The optician calls this "drying up each wet." 
This probably causes the surface of the tool to become quite 
warm, allowing the pitch surface to flow rather rapidly and 
to renew itself. 

Large lenses and very soft materials are best polished by 
coating the surface of the polisher at regular periods with 
fresh pitch or beeswax. The polishing tool is to be coated 
at intervals of 1 to 3 hours. The beeswax is applied to the 
facets of the tool with a swab made of cheesecloth bound on a 
short stick. It is advisable to have the wax smoking hot 
and to apply as thin a coating as possible. In polishing 
speculum metal, which scratches rather easily, the fresh 
beeswax coating is to be charged with dry rouge. The 
rouge is applied to the facets with the tip of the finger. 


When a full polish is achieved, that is, when the grinding 
pits are entirely removed, the work is ready for testing and 
figuring. A convenient and simple test for full polish is to 
focus sunlight on the glass surface with a lens. The focus 
of this lens does not heat the glass much, but light scattered 
by pits in the surface is quite conspicuous if the surface is 
not fully polished. 

To avoid introducing astigmatism into the work during 
polishing, it is frequently rotated a fraction of a revolution 
with respect to the supporting table in order to distribute 
the effect of edge arcs symmetrically around its periphery. 

Figuring. Figuring is the process whereby a polished 
surface has its shape altered by local working with polishing 
tools. For example, a spherical surface is made aspheric, or 
undesirable zones or astigmatism is removed. 

Sometimes in figuring plane parallels or prisms the effect 
of inhomogeneities in physical properties of the glass can be 
corrected (in first approximation) by slight deviations from 
flatness in the surfaces. 

The general procedure in figuring is one of trial and error. 
Testing is alternated with local polishing on those areas 
which are high in reference to a desired surface. 

Cutting zones and transition zones. The behavior of the 
polishing tool depends on its size, character of faceting, shape, 
and the manner in which it is manipulated on the work. 
There is no way in which a tool may be manipulated so that 
it will remove glass from a surface uniformly. Rather, each 
manipulation, if carried out on a perfectly flat surface, 
tends to produce its own characteristic zones, which will be 
referred to as the cutting zones of the tool. Figs. 17 and 
18 illustrate the cutting zones of some typical tools. These 
zones are defects in the mirror surface symmetrically posi- 
tioned about the center of the work. The figuring proce- 
dure consists in testing the imperfect surface and working 
it with a suitable tool whose cutting zones will tend to 
cancel the zones revealed by the test. 

Sharp zones are first "softened" with a large tool coated 


with soft pitch. This procedure applies both to those zones 
remaining from polishing and to those which may appear 
during the figuring. The latter are usually transition zones 
resulting from imperfect cancellation of a smooth zone in the 
work by the cutting zone of the tool. This is illustrated 
by Fig. 20. After the sharp zones are softened with a soft 
pitch tool, the optician tests again to determine the figure. 
To carry the figuring farther, a satisfactory surface tangent 
to the " valleys" of the surface, lying wholly within the glass, 
is imagined, and the hills relative to this imagined surface 
are polished away with an appropriate tool and stroke. 
This cycle of testing, polishing in a manner such that the 
cutting zones improve the figure, testing, smoothing transi- 
tion zones with a soft tool, testing, and so forth, is continued, 
until the necessary figure is attained. 

Interpretation of the action of polishing and figuring tools. 
If we could assign quantitative values to all of the factors 
influencing the cutting action of any given tool and stroke, 
we could conceivably predict the cutting zone for it. How- 
ever, we cannot do this; but we can describe the factors 
qualitatively as they are appraised in the minds of opticians. 

First, the polishing tool cuts away the glass in proportion 
to the time the tool is passing over the glass. 

Second, the tool cuts faster as the speed increases. The 
cutting or polishing rate is not, however, proportional to the 
speed at which the tool passes over the work. 

Third, sections of the tool which overhang the work dur- 
ing a part of the stroke cut relatively faster than the sections 
which do not overhang the work. 

Fourth, the facets of the tool which lead cut faster than 
following facets, because new rouge available to the leading 
facets is wiped away from the path of following facets. 

Fifth, the tool cuts fastest where the pressure on it is 
greatest, everything else being equal. This accounts for 
the selective action of the full-size tool on high zones, which 
action is the basis of all figuring. It is important to give 
this factor careful consideration in working aspheric surfaces, 


in which the tool naturally works in a way that tends to 
return the surface to a sphere. 

Figuring tools for zones. Figs. 17 and 18 show various- 
shaped tools and illustrate the zones which they would 
ordinarily produce in a true flat surface when worked with 
both long and short strokes. The stroke in each case is a 
thin oval across the center of the work. The use of an oval 
stroke has an advantage over a straight reciprocating stroke 
in that the tool never comes to a complete stop. 

It will be noted from Fig. 17 that the full-size tool makes 
the work more convex by an amount which increases with 
the length of the stroke. Intermediate-sized tools, as the 
§ size, hardly change the over-all curvature of the work 
when a long stroke is used, while a short stroke with this tool 
makes the work more concave. Smaller tools make the 
work more concave. 

It will be further noted that the effect of the tool in chang- 
ing the over-all curvature is (except for the case noted) 
greater than its effect in producing cutting zones. This 
change of curvature is generally inconsequential, except 
where one is making flats or striving for a radius of curva- 
ture specified to extreme precision. 

Fig. 18 shows the action of ring and star polishing tools. 

The behavior of tools on short radius curves may differ 
considerably from their behavior on flats. Figs. 17 and 18 
refer to flats. 

As testing methods are not very precise and the polishing 
methods even less so, it is well to approach the desired sur- 
face carefully and slowly, with periods of polishing inter- 
rupted frequently for testing. This allows one continually 
to change the "stratagem according to the tactical situation 
and nature of the terrain." 

One should use a clock to time the work done with a given 
polishing tool. If a mirror is improved by a certain treat- 
ment of 20 minutes' duration and the test shows that about 
as much more work is required, it is advisable to continue 
the treatment for 10 or 15 minutes more and test again in 






full-size hard tool 


full-size soft -tool 


%-size hard tool 


short stroke 


long stroke 



short stroke 

long stroke 

long stroke 

stroke from center to rim 



Y2.-s\z.e tool stroke across the center 

Fig. 17. Cutting zones for various tools. 

Chap. II] 



order not to overreach the desired result. It must also be 
emphasized that tools may cut faster at first than later, so 
that the significance of the time factor should not be taken 
too seriously. Also, the behavior of any given tool may be 
erratic. It is best to try it for short periods at a time with 
frequent testing in order to be certain of its action. Inas- 

star tool We size 

ring tool 2/3 size 

hollow-star tool 2/3 size 

j +-V pi 


polygonal -ringr tool s /s size 

Fig. 18. Cutting zones for various tools. 



A handle 
from an old 
may be 
used to 


much as the figuring procedure should not be hurried, 
beeswax-coated tools, which polish about three times as fast 
as uncoated tools, are not used for figuring. During the final 
stages of figuring, when delicate testing is required, the work 
should be allowed to stand on the testing support for suffi- 
cient time to allow complete 
equalization of temperatures. 
Pressing the tool through 
coarse cloth (such as an onion 
sack) gives many small facets 
in addition to the large facets. 
This results in quick contact 
of the tool to the glass and 
smooth action of the figuring 
tool from the start. (See 
Fig. 19.) 

To avoid astigmatism, the 
work should be occasionally 
rotated on the supporting 
table. In addition, with 
small tools it is important 
always to work the tool 
around the optical surface 
Fig. 19. through an integral number 

of revolutions. 
Hard tools tend to maintain a surface spherical or flat 
and are useful for generating flats or mirrors which are being 
worked to a specified radius. On the other hand, soft tools 
are recommended for working aspheric surfaces. Mirrors 
made by amateurs may exhibit a better figure than mirrors 
turned out by professional workers. The reason for this 
lies in the fact that amateurs usually use soft tools, which 
produce smooth flowing zones. On the other hand, pro- 
fessional opticians have the skill and knowledge to remove 
zones quickly with harder tools. In many cases, this rapid 
working produces faint transition zones, which show up 
under the most severe testing conditions. It is character- 

A piece of '•'♦^Marquisette'' 
wet with soap and glycerin, 
and placed between the 
worK and the tool will 
divide the pitch facets 
into tiny facets about Vs 
inch sojuare. 



istic of the commercial optician that he will produce a figure 
as good as, but no better than, that which his specifications 
call for. 

Manner of figuring various zonal defects, and of making 
aspheric surfaces of revolution. Focograms and exaggerated 
profile curves illustrating the manner of figuring various 
symmetrical defects are shown in Figs. 20, 21, 22, and 23. 

fMmmmmk ^ 




turned -down edge 
ttnife-edgfe test 

center cut away 
zones introduced 

treatment for 
a turned-down 


soft J%£$ £T^ 



Fig. 20. 

zones removed 

The interpretation of focograms is described in a later para- 
graph. At the upper left of Fig. 20 we see the focogram 
and exaggerated profile of a mirror with turned-down edge. 
This is corrected as follows : A f -size tool and short stroke is 
used. Two cutting zones are produced. One zone is posi- 
tioned where the leading edge of the tool comes to the ex- 
treme limit of the stroke, and the other is positioned where 
the trailing edge of the tool comes to the limit of the stroke. 
Besides making cutting zones, the tool has the further effect 


of making the figure more concave. The result is to change 
the full-line profile curve at the top left of the figure or the 
dotted profile curve at the top center to the profile exhibiting 
two sharp transition zones, as shown by the full curve at the 
top center and right. These transition zones are then 
smoothed off with a full-size soft tool with relieved edges 
(to avoid a turned-down edge), using a short stroke. 

The treatment with a f-size tool, as described above, is 
suitable for removing a turned-down edge from a circular 
flat; since the final treatment with a soft tool makes the work 
more convex (see Fig. 17), one can, by the judicious balance 
of the work done with the two tools, balance the increase in 
concavity produced by the first by the increase in convexity 
produced by the second. Fig. 20 shows the procedure 
applied to a spherical surface. 

The sketches at the top of Fig. 21 show how a turned-up 
edge is turned down with a full-size soft tool. 

The second series of sketches of Fig. 21 shows two methods 
of figuring to remove an intermediary depressed zone. The 
profile of the full line at the left or the dotted line in the 
center is changed by the indicated treatment, shown in the 
center, to the full-line profile in the center or the dotted-line 
profile at the right. In turn, this is changed by the indicated 
treatment to the spherical curve represented by the full-line 
profile at the right. In the first treatment the existing 
dotted profile, center, is elevated at the center and has a 
turned-up edge in reference to the imagined curve repre- 
sented by the full-line profile. This imagined curve is 
realized with a sub-diameter tool. At the right the full line 
represents the imagined surface which is realized by remov- 
ing the narrow sharp transition zones with a full-size soft 
tool. Inasmuch as this treatment does not change the 
radius of the work, it is suitable for figuring flats. 

By the alternate treatment, which decreases the concavity 
of the mirror, the cutting zones of the soft full-size tool 
change the intermediary depressed zone (depressed in 
reference to an imagined spherical surface) to two sharp 

Chap. II] 


turned-up edge 






SOft JS3J 

tool +~ 



(I fill 13 El III 



toS -*■■**- 

radius unchanged 





alternate treatment- radius lengthened 

', or de- g|jj|g||?gg|£p 
pressed tygggggp 

■treatment -radius unchanged 

alternate treatment- radius shortened 

Fig. 21. 



[Chap. II 

elevated zones on a second imagined spherical surface. 
These elevated zones are then managed with a sub-diameter 
soft tool as illustrated. 

Two treatments for a small depressed zone near the center 
of the work are illustrated in the bottom series of Fig. 21. 

r~ — n 


centerj^#- lUi 

facet 1EI13E3E1 




r 1 


Fig. 22. 


r — -i 


By one, the first imagined surface lying wholly under the 
glass surface requires the removal of an outer layer repre- 
sented by the difference between the dotted starting profile 
in the center and the full-line final profile in the center. 
The next imagined surface, now a spheric one, leaves several 


sharp zones to be removed by the usual treatment with a 
full-size soft tool. 

The alternate treatment goes from the primary defective 
surface to one with an intermediary elevated zone relative to 
the desired spheric surface. A chordal stroke is used. The 
elevated zone is removed by a second larger-size tool. 

In working on small zones in large mirrors or relatively 
large zones in small mirrors the optician has as possible 
figuring tools the thumb, the fingers, and the ball and heel 
of his hand. 

Fig. 22 shows how a narrow elevated zone may be removed 
with the thumb and how a depression may be removed with 
tools from which a facet has been removed. The thumb is 
used with extreme caution, applied lightly for one revolution 
at first, and then, if necessary, for a few mure complete 
revolutions. There is a danger of overcompensating for the 
elevated zone with the cutting zone of the thumb, because 
small polishing tools cut very rapidly. 

In testing an optical surface which has been figured by 
the fingers, one must allow enough time for the heat de- 
veloped by the friction of the fingers to be dissipated. 
Even for one revolution, this heating will produce a false 
zone, by expansion of the glass, which may be higher than 
the original zone. 

The top of Fig. 23 shows how a spheric mirror can be 
parabolized with a star tool. The focogram at the top right 
gives the appearance of the parabolic surface when it is 
tested at the center of curvature. The focogram of a para- 
bolic profile tested at the mean center of curvature exhibits 
the character of a soft raised intermediary zone. 

The second series of Fig. 23 shows an alternate parabol- 
izing procedure and f ocograms of the appearance of the figure 
of the mirror, as tested at the focus, before and after the use 
of the method. The advantage of testing a parabolic 
mirror at the focus is evident : The optician works toward a 
uniform distribution of light over the mirror face. The ad- 
vantage of testing at the focus over testing at the center of 



[Chap. II 

spherical mirror 
tested at the 
center of 

method of correcting, 
focal length decreases 

mirror tested at 
the average 
center of 

spherical mirror 
tested at the 
focus, i.e., with 

an optical flat 

alternate method of 
correcting, focal 
length increases 

mirror tested 
at the focus, 
i.e., with an 
optical flat 

spherical convex 
mirror tested 
with a spherical 
mirror or with a 
paraboloid and flat 

method of correcting 

spherical lens 
tested with an 
optical flat 

method of correcting 

Fig. 23. 

hyperbolic con- 
vex mirror 
tested with a 
spherical mirror, 
or paraboloid 
ctnd flat 

hyperbolic lens 
tested with 
an optical flat 


curvature is especially great when zones are being removed. 
Zones which are practically invisible if the mirror is ex- 
amined at the center of curvature become quite conspicuous 
when it is tested at the focus. 

The last two series of Fig. 23 show procedures that may be 
used for hyperbolizing. 

Astigmatism. The correction of astigmatism is more 
difficult than the removal of central symmetric zones. Cyl- 
indrical defects and, in general, all defects which are not 
symmetrical about the center of the work produce astig- 
matism. These defects must be worked out by hand. The 
rule of procedure is the same as it is for zonal defects — namely, 
the polishing is done on the high portions of the surface. 
Transition zones are removed with a full-size tool in the 
regular manner. The simplicity of this rule must not, 
however, be allowed to obscure the fact that the correction 
of astigmatism is one of the most delicate operations re- 
quired of the optician, and that aside from a knowledge of 
what is to be done it requires considerable manual dexterity. 
The tendency of tools to cut fastest near their periphery 
and especially where their edges come to rest is to be con- 
tinually kept in mind. The complete removal of astigma- 
tism in an optical surface is the apogee of good workmanship, 
while its avoidance is the result of experience. 

Optical testing. There are many applications for optical 
tests besides their employment to guide figuring. For 
example, one may wish to know the figure of a finished spheri- 
cal concave mirror, a flat, or perhaps a lens of unknown 
quality. Also, the testing methods described here can be 
used to test gratings. The Foucault knife-edge test is em- 
ployed by the Schlieren-methode for photography of sound 
wave fronts. 6 

Newton's fringes. The simplest optical tests are inter- 
ference tests using monochromatic light of wave length A. 

«T6pler, A., Pogg. Ann.,. 131, 33, 180 (1867). 

Wood, R. W., Physical Optics, page 93. New York: The Macmillan 
Company, 1934. 


The fringes manifest by a thin air film between optical 
surfaces are called Newton's fringes. They represent lines 
of equal optical separation of the surfaces. Between two 
adjacent fringes the optical thickness of the air film varies 
by an amount A/2, and the fringes may be interpreted as 
contour lines for the surface of one glass referred to the sur- 
face of the other glass, which is usually a flat or spherical 
test surface. 

The fringe system between two flats, if they are slightly 
inclined to each other and are illuminated with monochro- 
matic light, is a series of parallel equispaced straight lines. 

A cylindric surface of long radius of curvature in contact 
on a line with a flat gives straight fringes of unequal spacing. 

A spherical convex or concave on a flat gives concentric 

Fig. 24 illustrates a box for testing optical surfaces in 
contact with a flat and the appearance of the fringes under 
different conditions. 

The appearance of a convex spheric or cylindric surface 
on a flat is the same as the appearance of a concave surface. 
The difference in distance between the surfaces at one fringe 
and at the adjacent fringe is A/2; but the sign of the differ- 
ence, that is, whether the separation is increasing or de- 
creasing, is not known. The following rule may be employed 
to differentiate between a concave and a convex surface. 
The exhibited curved fringes expand away from their center 
of curvature when the head is lowered or moved away from 
the normal to the flat if the surface is convex, while, if it is 
concave, they contract toward their center of curvature. 

Newton's fringes are particularly suited to making a com- 
parison between a "flat" of unknown quality and a master 
flat. Also, they may be employed for testing surfaces of a 
definite radius by pressing them against a master plate of 
the same radius but opposite curvature. In this case, white 
light rather than monochromatic light is generally used, and 
deviations from the master are determined by the residual 
color of the interference pattern. 



angle of vision 
for observing 

screen of white 
"flashed" glass 
or ground glass * 

optical flat 

and work being 


apparatus for 

observing Newton's fringes 

provide vent holes 
in the back 

space for 

mercury arc, 
neon, helium 
or other 
discharge tube 

light source 

Newton's fringes between "two flats 

inclined slightly inclined more greatly inclined 

Newton's fringes between a flat; and a cylinder 2. A concave 

surfaces tilted 
along the axis 
of the cylinder 


surfaces parallel 
Newton's fringes between aflat and a sphere one A. convex 

surfaces parallel 


inclined slightly 

Fig. 24. 

greatly inclined 



When precision tests are made of a "flat" against a master 
flat and the fringes are observed at other than normal 
incidence, it is necessary to have the fringes running parallel 
to the plane of reflection, or they will be curved, even though 
the work is flat. Deviations of the fringes from a straight 
line are estimated by comparison with a stretched wire or 

Haidinger's fringes. Haidinger's fringes are excellent for 
testing the quality of plane parallels. Fig. 25 shows how 

■ discharge tube 


fa >«discnarae xudc . macK 

plane \jL ^ "g"* source /background 

parallel " 


plane parallel 
being tested 

diffusing screen 

discharge tube 
light source 


This arrangement 
is best if the plane 
parallel is half- 
silvered on both 

This arrangement 
will do if the plane 
parallel is half- 
silvered or not. 

reflection of the 
observer's e>e 
The appearance of 
Haidinger's fringes 

Fig. 25. 

Haidinger's fringes are observed. The usual manner of 
observing Haidinger's fringes is shown at the right in Fig. 25, 
and the appearance of the fringes and their positioning in 
respect to the reflected image of the observer's eye are shown 
in the lower right of this figure. These fringes are arranged 
like the Newton's fringes produced by a sphere on a plane. 
They are different in that Newton's fringes are observed by 
focusing the eye on the thin air film between the plane and 
sphere, whereas Haidinger's fringes are observed at infinity 
either with the eye or with a telescope. Newton's fringes 


represent the locus of points of equal optical thickness, while 
Haidinger's fringes represent the locus of points where 
rays "from the eye" make equal inclination to the plane- 
parallel plate. The Haidinger fringes are observed at nor- 
mal incidence, and the plane-parallel plate is moved laterally 
to apply the test to different areas of the work. A variation 
of thickness, from one end of a plane parallel to the other, 
results in the appearance or disappearance of fringes. If 
the plate gets thicker, fringes appear. The appearance and 
disappearance of one ring corresponds to \/2n change in 
thickness. For the most delicate testing, a telescope is 
used. A large field telescope equipped with a filar mi- 
crometer may be used to measure the diameter of the rings. 
With this telescope the appearance of about one tenth of a 
ring can be detected. For glass having an index of 1.5, 
one tenth of a ring represents a difference of thickness of 
1.5 X 10~ 6 cm or approximately 0.5 X 10 -6 inch. 

Eyepiece tests. Another important means of examining 
the quality of image-forming systems of mirrors or lenses 
is to inspect an imaged pinhole light source. A high-power 
magnifier, such as a 14X Hastings triplet, is suitable for 
examining the image. This test is called the eyepiece test 
because it is essentially the test which is applied when one 
observes a star in the eyepiece of an astronomical telescope 
on a good night. The eyepiece test is the most sensitive 
optical test for astigmatism. The infrafocal and extra- 
focal images should be examined, as well as the focal image. 
It is advisable to record the results of this test by drawing 
rough distribution curves representing the light intensity 
along a horizontal diameter of the image. Fig. 26(a) 
illustrates eyepiece images for a good (though not perfect) 
spheric mirror tested at the center of curvature. Fig. 26(b) 
illustrates the eyepiece test for an overcorrected parabolic 
mirror tested at the center of curvature, and (c) illustrates 
the test applied to a mirror with very slight astigmatism. 

Fig. 33 illustrates the eyepiece test in comparison with the 
Foucault, Ronchi, and Hartmann tests. 



This was a very good spherical mirror 
10 inches in diameter, f/w It had a depressed 
centra) zone, ^/^o deep, barely detectable 
with the Foucault Test. 

appearance of the image of an artificial 
star (pinhole ,ooos"in diam) as seen with 
a 2o-power eyepiece 

/^" -0.05" 

/ inside focus 

light-intensity curves 

at foe us 

+ 0.05" +0.10" (a) 

outside focus 

This was an overcorrected parabolic 
mirror 12 inches in diameter, yb. The 
test was made at the center of curvature 
(f/i6) and it therefore appeared to have 
a turned-down edge and a depressed center. 

appearance of the image of an artificial 
stat* ( pinhole and eyepiece same as above) 

-aio" -o.o5" o.oo" 

inside focus at focus of "a" 

light -intensity curves 

40.05" +O.10" (b) 

outside focus 

This is a typical astigmatic mirror. 
Astigmatism is only detectable in the 
Foucault Test when it is very marked. 

appearance of the imagfe of an 
artificial star 


focus in 

circle of 

focus in 



outside (c) 

Fig. 26. Eyepiece image test. 

Chap. II] 


Foucault knife-edge test. The Foucault knife-edge test is 
usually employed for the detection of central spherical aber- 
ration, particularly in testing work of fairly large aperture, 
such as lenses or mirrors used for astronomical telescopes. 

The test is simplest as applied to a spherical concave 
surface of long radius. A small hole is pierced in thin metal 



Knife edge 
inside focus 

knife edge 
at focus 

Knife edge 
outside focus 

Fig. 27. 

Knife edge 
inside focus of b 
at focus of a 
outside focus of c 

sheet with a sharp needle and illuminated by means of a 
lamp, together with a suitable optical system. 7 This hole 
is located near the center of curvature of the mirror. (See 
Fig. 27.) Light from it is reflected by the mirror to form an 
image at an equal distance on the opposite side of the center 

7 Several layers of the thin metal sheet are laid together on an anvil, and a 
sharp needle is driven halfway through them. They are then separated and 
the one with a suitable hole is. selected. Each pierced sheet has a small hole of 
a different size and all the holes are round. 

Knife edge 

Spherical mirror 

f \ Knife 

'good microscope edge 

/ja x Spherical mirror-short focus 

Ca) (b) 

test mirror 

Flat mirror- 

Parabolic mirror 


star light a 

flat test / 
mirror / 

Parabolic mirror -focal length 
, . of zone *a" — *1+Jb »f Knife 
(Cy edge and pinhotevSriove together 

"-■IZ. Q-^flat 
~~~^ test 


(c.) Parabolic mirror 


flat test mirror spherical test mirror 

Parabolic mirror (perforated) Hyperbolic mirror (convex) 

(g) Cassegrain secondary (h) 

/V'-*-4 ^silvered bead 
~ — — " — "\ /&' light focused on bead 

mirror test mirror Ellipsoidal mirror .. 

Hyperbolic mirror (convex) (i) Gregorian secondary (j) 



focal ten 
•j>' of zone 

if knife 

edge and 

mov ?L f&&-^ light focused VI/ 

together *^> filter & on silvered bead 
(k) Schmidt Camera correcting plate 

spherical test mirror 

5chrnidt Camera 
correcting plate 

Fig. 28. Arrangements for carrying out the Foucault test: 

(a) Testing of a spherical mirror at the center of curvature. 

(b) Testing of a short-focus spherical mirror. 

(c) Testing a flat with an auxiliary spherical mirror. If knife edge and 
pinhole move together, a lack of flatness represented by a sagittal 
distance h introduces astigmatism between the horizontal and vertical 
focus. The relation between h and A R is given. 



(d) Zonal testing of a parabolic mirror at the center of curvature. R is 
the radius of curvature at the center. 

(e) Testing of a parabolic mirror at the focus with starlight. 

(f) Testing of a parabolic mirror at the focus with an auxiliary testing flat. 

(g) Alternate procedure for testing a parabolic mirror with a flat, 
(h) Testing of an hyperbolic mirror with a spherical testing mirror, 
(i) Testing of an hyperbolic mirror with a parabolic mirror and flat, 
(j) Testing of an elliptic mirror. 

(k) and (1) Testing of a Schmidt lens. 

of curvature. When the eye is placed behind this image so 
as to receive light from all parts of the mirror, the whole 
aperture will appear evenly illuminated. Then if an opaque 
screen, the so-called knife edge, is moved laterally across the 
focal point, the whole aperture will appear to darken evenly, 
that is, if the mirror is truly spherical, as in Fig. 27(h). 
If the knife edge is moved across the cone of light a short 
distance inside the focus, its shadow, as it appears on the 
mirror, moves in the same direction as the knife edge, 
Fig. 27 (i); if it is placed outside the focus, the shadow moves 
in the opposite direction, Fig. 27 (hi). It is possible by this 
means to locate the focus of the mirror with great precision. 

In the case of an imperfect mirror, such as the one shown 
in Fig. 27 (iv), all rays do not converge to a single point, and 
if the knife edge intercepts the converging light rays as re- 
flected from the mirror, the aperture will appear to be un- 
equally illuminated; some rays are completely cut off by 
the knife edge, whereas others pass by it and so reach the 
eye. The mirror shown at the right has an intermediary 
raised zone. For the inside half of the zone (c) the focal 
length is shorter than it is for the outside half (b). "With the 
knife edge advanced into the mean focus of the converging 
rays, those rays from the areas which have their center of 
curvature exactly at the position of the knife edge are 
attenuated; rays from (c) reach the eye without attenuation, 
while rays from (b) are cut off entirely by the knife edge. 
Accordingly, (c) and (b) appear very bright and dark re- 

The appearance of the mirror with the knife edge and eye 
in the positions indicated is as if it were made of plaster and 



illuminated at grazing incidence with an imaginary light 
source. 8 Usually the pinhole source is on the right side of 
the center of curvature and the eye on the left. In this 
case, if the knife edge cuts the image from left to right, the 
observer, thinking of this imaginary light source as illumi- 
nating the plaster disk from the right, interprets its shadows 
in accordance with their apparent inclination. For a lens, 
the observer interprets the shadows by thinking of the il- 
lumination as coming from the left. 

Various more complicated setups for making the Foucault 
test are shown in Figs. 28 and 29. In these two figures a 

^j— „ . monochromatic 


light source 

Lena for use at conjugate foci 

>sf I at test mirror- 
Telescope Lens 

color/ / t- ,n y ^perfection 
filter 4 in a silvered mirror 
condensing lenses 

Microscope objective 

Fig. 29. Foucault tests for achromats to be used to focus parallel light 
and diverging light. Also, tests for a microscope objective and a prism. 


flattest mim>r c Vr?£ , / tic 
— *. light 


test mirror or lens of unquestioned quality (or at least of 
known quality) is shown clear, while the tested mirror or 
lens is cross-hatched. 

Zonal knife-edge testing. Aspherical mirrors, such as 
paraboloidal ones, can be tested at their mean center of 
curvature without an auxiliary testing flat by measuring the 
radii of curvature of the glass at various zones. The mirror 
is covered with a diaphragm of cardboard with holes opposite 
the zones to be tested. The simplest diaphragm has holes 
at the center, at the edge, and at 0.707 of the radius, as 

8 See article on Foucault's shadows by E. Gaviola, in Amateur Telescope 
Making, Advanced (Albert G. Ingalls, editor), page 76. New York: Scien- 
tific American Publishing Company, 1937. 



cardboard diaphra; 
same size a 

illustrated in Fig. 30. If the mirror is parabolic, the char- 
acteristic shadow shown in the upper right of Fig. 23 will 
appear when the mirror is viewed without the diaphragm. 
With the diaphragm, the measured difference in focus of the 
center and the zone at the rim 
will be r 2 /2R, where R is the 
mean radius of curvature of 
the mirror and r the radius of 
the mirror. This test is suit- 
able for testing small mirrors 
to determine when a raised 
intermediary zone, as illus- 
trated in Fig. 23, is carried 
sufficiently far to parabolize 
the mirror. 

Diaphragms for very large 
parabolic mirrors, or mirrors 
of focal ratio of //4.5 or 
greater, are constructed so 
that the radius of curvature 
of a large number of zones 
can be measured. Ordinarily, 
mirrors of aperture //10 or 
less need not be parabolized 
unless their diameter is 
greater than 24 inches. 

Another procedure for 

Typical zonal diaphragm of 
a "parabolic mirror 

for this diaphragm, if the 
focal length of the center 
is R, the focal length of 
the intermediate zone should 
beK+ 8k,and the focal 
length of the rim should 
be^t^ R . Example — 
for a 6-inch mirror, f = 4 
feet , the focus at the 
center will be 96 inches, at 
the intermediate zone 
96"+ 0.023", and at the 
rim 96" -r 0.047." The 
dimension R (96) need not 
be precisely measured. 

quantitative application of 

the knife-edge setup has been described by E. Gaviola. 9 By 
this procedure the inclination of different zones is determined 
relative to a mean surface for the mirror by measuring the 
position of the knife edge which intersects the light rays 
reflected from these zones. 

In any knife-edge setup, and especially where accurate 
quantitative zonal measurements are to be made, it is im- 
portant to avoid parallax. Although a setup like the one 

9 Gaviola, E., J.O.S.A., 26, 163 (1936). 



detail of 

90° prism - 7 /zz face 
silvered and 

A>* square-hole 
JgT knife edge 

handle for 
.^ rotating 

f*- x A-*j a pinhole in shim 
vertical slide for stock is clipped 
adjusting height under this spring 

of pinhole « 

mirror being* 



square hole , the'' 
sides of which are 
used as knife edges < -\- > 

location of observers eye-* 
tangent screw 
■for making s 
vertical cuPoff 

ball keep 


ball bearing 

half nut mounted on 
a spring so it can be 

ised f 


or quick adjustments 

screw for measuring 
focal lengths of 
of different zones *~~ 

calibration in 
of an inch 

K M \/ 


( Bausch 6*Lomb 
"sport glass" lenses 
separated ZYY 


observers eye ScaJc of , nchcs 

vent holes in lamp 
housing top and 


G.E. projection lamp 
6v-9a with ribbon 
filament loxi.smm 

Fig. 31. 



shown in Fig. 27, using a simple pinhole and knife, is suitable 
for qualitative tests on small mirrors of long radius, a more 
elaborate setup is usually required. 

Parallax appears whenever a Foucault test is made on 
more than one mirror, such as a test on a parabolic mirror 
at its focus with an auxiliary flat. (See Fig. 28.) This 
parallax is due to the fact that a ray from the pinhole 
strikes the parabolic mirror at one point and is subsequently 
reflected, after returning from the flat, at a different point on 
the parabolic mirror. The displacement between these 
points on the paraboloid mirror is somewhat less than the 
displacement between the pinhole and the knife edge. 
Even so, the displacement may be sufficient to yield results 
that are quite misleading. As a result of parallax one does 
not get an indication of the character of the mirror at either 
of the two indicated points but rather a kind of average for 
the two points. This information is of little value and may 
be quite misleading if the effect of errors at one point of the 
parabolic mirror is compensated by opposite errors at the 

A knife-edge testing device is illustrated in Fig. 31. 
Fig. 32 shows two attachments for it which may be used to 

to and from 
mirror ^ 

si I vcred v A 
pellicle ^ffl 
mirror v 

Knife edgre 

position of ^ 
observer's eye 


the knife 
edge forms 
one jaw 
of a 

from light 

position of 
observer's eye 


Fig. 32. 

avoid parallax. The device at the left eliminates parallax 
by eliminating the displacement between the knife edge and 
the pinhole — a virtual image of the pinhole formed by means 
of a half -silvered pellicle mirror lies exactly on the knife 

f*-about 20/2 

Schematic diagram of the focal 
rays of the mirror used in this il lustration. 
The actual mirror was 3 inches in diameter 
with a radius of curvature of 21 inches. 

Knife edge at 
focus of 2 

Knife edge at 
focus of y 

knife edgfe at 
focus or x 

Foucault Knife-Edge Test (Knife edge at the bottom) 

shield placed over 
mirror* to help 

identify zones 
Eyepiece Test 

The screen used had 
ISO lines per inch, 

eve piece focused 
at a — inside focus 

eyepiece focused 
at b -outsidefocus 

appearance of extra focal images 

I^pnchi Test screen inside focus(c) outside focus (d) 


r m ctal nnH^^Bi H-y 

over iH^^^IH I Uj I 

mirror ■■■■■■■ ^^^^mi^mm^z 

Hartmani) eyepiece inside focus (c) outside focus (f) 

Fig. 33. Comparison of Foucault, eyepiece, Ronchi, and Hartmanv 

tests of a defective mirror. 



edge. A pellicle mirror is made by flowing lacquer on an 
inclined glass plate. After the lacquer is hard, it is stripped 
off the glass under water. The film of lacquer is then 
mounted on the flat surface of a brass frame and half- 
silvered by the evaporation process. 

The device at the right uses a slit instead of a pinhole. 
This device avoids parallax by the simple expedient of remov- 
ing it by half a revolution in azimuth from the testing di- 

The Ronchi test. The Ronchi test will be treated here 
only briefly, since it is not widely employed and its inter- 
pretation is not simple. 10 A small light source illuminates 
the mirror through a ruled surface (about 100 lines per inch), 
and the image is also formed through another section of the 
same ruled surface. For comparison, the Foucault test of a 
defective mirror is shown at the top of Fig. 33. Below, we 
have the results of the eyepiece test and Ronchi test of the 
same mirror, and, at the bottom, the Hartmann test is il- 
lustrated. Fig. 34 shows a simple quick way of testing a 
lens with the Ronchi screen. A good lens gives straight 
Ronchi lines. 

Hartmann's test. Hartmann's test is similar to the Ronchi 
test for a lens, illustrated in Fig. 34. A diaphragm, such as 
the one shown at the bottom left of Fig. 33, is prepared with 
several appropriately spaced holes. This diaphragm is 
placed directly in front of the mirror or lens. The Foucault 
test determines errors in the mirror by the lateral positioning, 
relative to their neighbors, of various rays as they pass 
through the focus; the Hartmann test determines the posi- 
tions of these rays, relative to their neighbors, at points either 
inside the focus at (e) or outside of the focus at (f). The 
relative position of the holes in the plate and the appearance 
of the rays, as observed with the eyepiece at (e) and (f), are 
shown in the figure. 11 

10 Anderson, J. A., and Porter, R. W., Astrophys. J., 70, 175 (1929). 

11 For further treatment of optical testing, see articles contained m Amateur 
Telescope Making, Advanced, and references cited therein. 



[Chap. II 

The advantages of the Hartmann test over the others 
illustrated in Fig. 33 are that it is not necessary to locate 
the mean focus, and that the results of the test are easily 
recorded photographically, both of which features make it 
particularly useful for figuring lenses for the ultraviolet 
region of the spectrum. 

Lining up a system of mirrors. In the more complicated 
testing setups shown in Figs. 28 and 29 it is often quite 

piece of engraver's 

screen with 80 to | cn a being 

120 lines per /^ i» tested 

inch ^ Jy^T-'y^'^ ' 

j/^ - - // \\ r light from 
'' ' ■' ' a distant 

street lamp 

appearance of a 
simple lens 

appearance of a. 
corrected lens. 

Fig. 34. 

difficult to get the mirrors or lenses lined up. The appear- 
ance of coma in the eyepiece image, however, can be used to 
advantage for this. The coma of a system not properly 
lined up is quite strong and indicates clearly in what direc- 
tion the mirrors are to be adjusted to get round images. 

Some opticians put two white threads at right angles to 
each other across the face of one of the mirrors. When these 
threads and all of the secondary images of them viewed from 
the focus appear symmetrical, the system is in alignment. 

Two methods of generating optical surfaces. As we have 
already pointed out, the optician's task is defined as the 
generation of accurate surfaces on mirrors, lenses, prisms, 
and so forth, which possess a high polish. This is ordinarily 
done by hand or with the modified Draper machine, as 



described before. Or it may be done with a high-speed 
hand-lever machine in the manner described below. The 
procedure with a modified Draper machine or by hand is 
slow, but it yields the most accurate results. The modified 

b rass waf er cemented 
to lens 

hinge joint 

clips should be 
soldered to the 
bottom of the cake 
tin to hold it in place 
and permit its 
removal for cleaning. 

12 3 

scale of inches 

ball bearings of 
good quality may 
be used- Bearings 
must be protected 
from abrasives. 

A chuck arbor having the same 
taper as the polishing spindle 
should be obtained to fit the 
worKers lathe. 

C. 35. 

Draper machine illustrated in Fig. 10 uses a single crank and 
allows for counterbalancing of the tool, automatic control 
of the tool, slow smooth stroking, and easy placement of 
the work. In contrast with this, the hand-lever machine 
shown in Fig. 35 features high speed and simplicity. The 



[Chap. II 

spindle is run at about 100 to 600 r.p.m. Naturally, the 
heat thus generated, as well as the high speeds of cutting, 
makes work of the highest precision impossible. 

Working optical surfaces on the hand-lever machine. 
The tool used with the hand-lever machine may be attached 
to a high-speed spindle as shown in Fig. 35, or the work may 
be attached with wax to the spindle and the tool applied 
above. In the former case, a socket for the pivot point of 

length of rod equal 

5 r 22iVi H f 5^ vatUre Pnck punch marks 
of desired die ^ ip| {r ross 5hde 

convex die 


The carriage is held 
against the rod while 
the tool is fed across 
the work. 

prick punch 
mark in 
tail stock 

The rod is placed 
here for cutting 
concave dies. 

Fig. 36. 

the hand lever is waxed to the work with a mixture of 2 
parts coal-tar pitch to 1 part sieved wood ashes. When the 
tool is applied above, the socket is turned in the back of it. 
The preliminary grinding may be accomplished on the 
hand-lever machine with a sub-diameter ring tool of iron, 
as with the Draper machine. When this ring tool is moved 
back and forth across the center of the spinning work with 
a short stroke so that there is no overhang of the tool, the 
surface is made concave. A long stroke with overhang 
gives a convex surface. The tool is pivoted and allowed to 


spin freely, and 90 Carborundum with water is fed on it to 
accomplish the grinding. The growth of the curve in the 
work is measured with templates. These are usually cut 
on a lathe from a thin sheet of brass or bronze. 

The final grinding is done on the hand-lever machine with 
a spherical brass tool of the same diameter as the work. 
The spherical tool is made as follows: A brass male and 
female part are turned on the lathe to the same curvature in 
the manner illustrated by Fig. 36. These are then lapped 
together with Carborundum to generate complementary 
spherical surfaces. If a lathe is not available, they may be 
separately ground with the hand-lever machine to an approx- 
imate fit with a third metal ring tool and then lapped to- 
gether. It is important to cut a cavity in the center of the 
one to be used as the grinding tool. The cavity should have 
a diameter of about one twentieth of the tool diameter. 

After the proper curve is approached in the work by grind- 
ing with the ring tool, the spherical brass tool is substituted, 
and the final grinding is carried out with 90, F, and 600 
Carborundum and 302J emery. During the grinding pro- 
cess the offset of the tool in respect to the work should never 
be so great that the tool and the work rotate in opposite 

Polishing is accomplished on the hand-lever machine with 
rouge on a pitch lap. The brass tool used for grinding may 
be warmed and coated with a layer of hard pitch or pure 
beeswax for polishing. While the wax is still warm, it is 
pressed to the proper shape with the already fine-ground 
work (wet with soap and glycerin solution) to give a layer of 
pitch about f to \ inch in thickness. As for grinding, this 
lap is cut away in the center to form a cavity of one twentieth 
of the tool diameter. Also, the pitch is cut to form annular 
grooves. These grooves facilitate contact between the pitch 
and the work. The pitch lap should be frequently trimmed. 

If the central cavity removed is too broad, the tendency 
is to polish the edges first, while if the central cavity is small 
and pitch is trimmed off the edges of the tool, it will polish 


the center first. Also, whether the polish progresses faster 
in the center of the work or near the edge depends on the 
offset. A little offset favors polishing the edges fast, and a 
big offset makes the progress of polishing greater in the 
center. One can easily keep track of the progress of the 
polishing by shining a strong light on the work and observing 
the "grayness" produced by the residual grinding pits. 

It is difficult to balance all these factors, and in practice 
one should observe how the polish progresses. If the polish 
is not progressing satisfactorily, the offset can be altered or 
the tool trimmed accordingly. 

Figuring aspheric lenses on the hand-lever machine is 
accomplished by polishing with sub-diameter tools and star 
or ring tools. 

Small lenses are aligned on the spindle by tilting them 
while the blocking wax used to cement them in place is still 
warm. The spindle is turned slowly, and if an object, 
preferably a small light source, seen reflected in the surfaces 
does not describe an eccentric circle as the lens rotates, 
the alignment is complete. 

After being centered on a brass tube mounted in the head- 
stock of the lathe, in the same manner as described above, 
the work is edged with an iron tool and grits. (See Fig. 6.) 

Relationship between two optical surfaces. Although 
we have emphasized the phases of procedure which are 
important in generating the optical surface, we have not 
dealt extensively with the orientation of that surface with 
respect to the general form of the work or with respect to 
other optical surfaces. These are matters usually managed 
in an obvious manner. However, in the construction of 
prisms, especially right-angle prisms and plane parallels, 
the manner of getting proper relationship between the two 
flat surfaces involved is not so obvious. A half -hour of 
polishing on the Draper machine or a few minutes of polish- 
ing on the spindle machine will usually put enough polish 
on ground surfaces to allow their relationship to be tested 
on a spectrometer table or with other optical tests. Right- 



When the blank for a prism is cut 
from a bloek of optical glass, the 
remaining scraps are cut up to 
be used in blocking the blank. 

The pieces are arranged on & 
flat surface around /^the 
blank to approximate/ a 
solid dish of glass. A/auarter 
inch spaces are left^oetween 
them. A Vie -inch flayer of 
wax is painted (( between 


'/is" cardboard mat 

Cotton is stuffed bet 
pieces to within a 
inch of their top 

A section of iron pipe is placed 
over, the arrangement and 
a X A \c\yer of plaster of Paris 
is poured between the pieces. 

een the 



Another ^-inch layer of 
plaster is poured 
between the pieces 

The iron rin^ and its contents 

are inverted and the exposed 

wax is scraped down about 

Vsz". The glass surfaces are 

now ready to be optically 

worked as a continuous disk. 

Because the plaster changes shape 

with drying, the fine grinding, polishing 

and figuring should be done the same day, if possible. 

Fig. 37. 


angle prisms are tested by employing their property of de- 
viating a light beam exactly 180°. This test is sensitive to 
about 1 minute of angle when it is made with the naked eye, 
and if it is made using a telescope equipped with a Gauss 
eyepiece, it is sensitive to about 1 second of angle. 

Plane parallels are ground to be flat and parallel to about 
20,000 of an inch. Good micrometers are used to test the 
glass for parallelism to this accuracy. The final optical 
precision is obtained by figuring. The test using Haidinger's 
fringes, described before, is used to guide the figuring. 
Plane parallels are usually made with circular faces to be 
cut up later into rectangles if necessary. 

Blocking. Inasmuch as round glass surfaces are more 
easily figured than square or rectangular ones, it is advis- 
able to mount a prism blank in a metal ring as shown in 
Fig. 37, together with auxiliary glasses having the same 
coefficient of expansion, to make up a circular array of glass 
surfaces. This circular array is held in the metal ring with 
plaster of Paris, and the ensemble is then worked as a single 
disk of glass. The parts may be immersed in a single thick 
layer of plaster, but it is best to imbed them in a double 
layer of plaster as shown in Fig. 37. A mixture of 3 parts 
plaster to 2 parts water is used. This gives an almost non- 
shrinking, although not very strong, cement. The work ma}^ 
be coated first with a thin layer of beeswax in cases where 
free lime or moisture in the plaster might attack the glass. 
After the plaster of Paris sets, its surface is shellacked to 
make it impervious to water. The grinding and figuring 
should be finished in one day. Otherwise, owing to " aging" 
of the plaster, the central and auxiliary surfaces will not 
maintain satisfactory alignment. 

Quartz and calcite. When optical surfaces are generated 
on crystals, it is frequently required to orient the surfaces 
precisely with respect to the crystal axes. Fig. 38 illustrates 
the manner in which the optical axis of quartz is precisely 
located. The crystal is cut at each end with the mud or 
diamond saw, the cuts being made roughly perpendicular to 

Chap. II] 



Nicol prism analyzer 

quartz crystal 
being tested „ 
The ends have 
been eot and 
ground with 
150 Carborundum 
and cover plates 
have been 
cemented on witr 

protractor for 
reading the 
required change 
in the crystal 
ends— The 
should be 
1.4 times 
the true 
angle of tilt, 
to allow for 
refraction at 
the end faee. 

polarizer — 
window glass 
black on the 
under side 

glass table on 
mounted on 
a metal ring 
which may 
be rotated 

spot on 
glass table 
to mark the 
optical axis 


light source 

of rinds 
• s du^to 


t J IM 111! II I I l« CrVSta * 

appearance of a £* owth 
typical specimen 
(from a photograph) 


nota tion of analyzer 

motion of 

left-hand right-hand 
quartz quartz 

Fig. 38. 



the optical axis. These parallel saw cuts are then ground 
with the abrasives to grade 150 Carborundum and "arti- 
ficially polished' ' by cementing cover glasses to the ground 


crystal to be cut 

dihedral angle 
115^ lO' 

hard-drawn crystal 
copper wire being cut 
in plaster 

Fig. 39. Calcite : 

(a) Orientation of axis. 

(b) Cuts for making Nicols. 

(c) and (d) Procedure for cutting calcite to make Nicols. 

ends with balsam. Polarized light is used to determine the 
optical axis in the manner illustrated in the figure. The 
system of rings noted through the analyzer will remain 
stationary when the crystal is rotated about a vertical axis 
if the cuts are perpendicular to the optical axis. If the 


rings " wobble" as the crystal is rotated, the axis of rotation 
is to be tilted by means of the gimbals provided (or with 
wedges) until the fringes are stationary during rotation. 
The ends are then recut, taking account of refraction (see 
note on the figure), and the plate re tested to check the result. 

When properly cut quartz crystal is rotated in a clockwise 
direction, the rings close in toward the center if the crystal is 
left-handed quartz and move out if it is right-handed quartz. 

An irregular piece of quartz can be roughly examined for 
striae by immersion in a tank filled with a solution composed 
of 80 per cent ethyl cinnamate and 20 per cent xylol (by 
volume). Iron oxide surface stains may be removed from 
the crystal by washing in oxalic acid solution. 

The orientation of the principal axis of calcite is shown at 
the top of Fig. 39. To cut calcine for making Nicols, the 
crystal is oriented and mounted in a wooden form having 
the cutting plane defined by a preliminary saw cut. The cut 
through the crystal is made by hand by sawing through both 
plaster and crystal with a hard-drawn copper wire mounted 
in a scroll-saw frame and charged with Carborundum. 

Optical working of crystals. Quartz is the optician's 
favorite medium. Both the fused and the crystal material 
are ground and polished by the same procedure as glass. 

Calcite crystals, especially large ones, are expensive, and 
in addition they are soft and easily fractured. Accordingly 
calcite is always worked by hand with very light pressures. 
All but the smaller Carborundum grains tend to produce 
fractures in calcite, and therefore the series F-400-600 
Carborundum and 302| and 303J emery is recommended for 
working it. If prisms with very thin edges are to be made, 
400 Carborundum is used as the coarsest grit. A beeswax- 
coated pitch lap is satisfactory for polishing and figuring 
calcite. For figuring, calcite should be blocked with calcite 
of the same crystal orientation. 

Rock salt is polished and figured on a hard-pitch tool 
pressed with a glass pressing lap. 12 The desired figure is 

u Brashear, John A., Proc. of Am. Assn. for Adv. of Science, 38, 166 (1885). 



[Chap. II 

obtained with an overcorrected pressing lap, as the figure 
obtained on this material is usually convex with respect to 
that established in the tool by the pressing lap. For ex- 
ample, in making flats, one would use a slightly concave 
pressing lap (two of an inch in a 4-inch disk or 60-foot 
radius of curvature). Rouge in saturated salt solution is 
used to start the polishing. Fig. 40 shows the arrangement 
of the pitch tool which is placed below the work, and a 
chamois skin used for drying the work. The work is rubbed 
against the lap until the rouge is almost dry. It is then 

wooden disk 
voch-salt covered with 

finished faces chamois 
protected with 

When -the lap 
is nearly dry 
the last stroKe 
is carried 
directly onto 
the chamois 

pitch lap 
fed with 
rouge and 
saturated salt water 

Fig. 40. 

kept moist with the breath for the final strokes and is slipped 
off the tool onto the chamois to be dried. This technique 
should be practiced on test pieces before big work is under- 
taken. The first face of a prism or lens is lacquered or 
waxed to prevent attacks by moisture while the second face 
is being worked. The pitch tool is coated with beeswax. 
Beeswax is useful for even softer materials than rock salt, 
such as potassium chloride, potassium bromide, and even 
potassium iodide. 

Polishing of metals. Perhaps the most important metal 
in optics is speculum metal. It is very hard, exhibits a 
conchoidal fracture like glass, and is worked by the same 
procedure, being ground with the same sequence of grits. 
The polishing tool should have narrow facets. It is often 


advantageous to do the final figuring of speculum surfaces 
with the metal face turned down to avoid scratching. 

Stellite is also worked like glass except that longer grinding 
periods (two to three times that for glass) are required. 
Ordinarily pitch polishing tools are used with rouge or 
chromium oxide as polishing agent. One should try to 
"hold" the figure from the grinding stages until polishing 
is completed. 

Hard steel is worked in the same manner as glass. 

In grinding soft steel, a still softer metal, such as copper 
or lead, is used as a grinding tool. 

Soft steel and hard-drawn copper are difficult to polish, 
but they can often be managed with a polishing tool coated 
with a mixture of paraffin and oxide of tin. As much oxide 
is added to the molten paraffin as it will take without crum- 
bling. This mixture is applied hot with a swab to the pitch 
polishing Jap. 

If a metal tends to etch or discolor during polishing, it is 
advisable to try carbon as a polishing agent (charcoal ground 
in a ball mill and washed). Chromium oxide will often give 
a bright polish in cases where rouge would discolor the metal. 

The very soft metals — silver, soft copper, gold, and so 
forth — cannot be easily surfaced by the ordinary optical 
methods. They become charged with the grits and refuse 
to grind. Silver circles are brightened by rubbing moistened 
cigarette ashes with the thumb back and forth in a direction 
parallel to the engraved lines. 

The Schmidt camera. The Schmidt camera is an image- 
forming device which combines features not possessed by any 
lens system, and, while it has some disadvantages, it may 
well prove to be a natural solution to many more instrumental 
problems than those to which it has already been applied. 13 

The camera has had considerable application in astron- 
omy, particularly meteor photography and survey work of 

13 Stromgren, B., "Das Schmidtsche Spiegelteleskop," Vierteljahrschrift der 
Astronomischen Gesellschaft, 70, 65 (1935). 
Smiley, C. H., "The Schmidt Camera." Popular Astronomy, U> 415 (19.36). 


large star fields. It has been applied to stellar spectrographs, 
and it is believed that it will have other applications in re- 
search where extreme speed, a long spectral range, and a large 
field are important. Two awkward features of the Schmidt 
camera are its curved focal plane and the inaccessibility of 
the plate or film holder. The curvature of the focal surface 
is R/2, where R is the radius of cuivature of the primary 
mirror. The focal surface is convex toward the sphere. 

The construction of a Schmidt camera is so difficult that 
it should not be undertaken except by one with considerable 
optical experience. The following is intended primarily as 
a description of it. 

The scheme of the instrument is shown at the top of 
Fig. 41. It consists of a spherical primary mirror and a 
Schmidt lens, which corrects the primary mirror for spherical 
aberration. The lens is located at the center of curvature of 
the spherical mirror, and its deviation from flatness is so small 
that no great error of achromatism is produced by the dis- 
persion of the index of refraction of the glass from which it is 
made. However, as an optical figuring job this deviation is 
great enough to make the construction of the lens difficult. 
This is because the curve deviates as much from any sphere 
as it does from flatness, so that all the construction difficul- 
ties of making aspheric surfaces are encountered. 

The Schmidt lens may have several contours, as illustrated 
in Fig. 41. The variation of thickness may be obtained by 
putting the curves entirely on one side of the plate or on 
both sides. The variation in the thickness At of a plate of 
diameter 2r, expressed as a function of the distance from 
the center of the plate y and the radius of curvature R of 
the primary spherical mirror, can be represented by any 

one of the family of curves 

_ y* - ky 2 r 2 
At ~ 4(n - 1)# 3 ' 

where ft may have any value between and 4. 

The characteristics of some of the curves are as follows: 
Where k = 4, the lens is too thick; where k = 0, the slope 

Chap. II] 



at the edge is so steep that the construction difficulties 
are great; where k = 1.5, the achromatism is best; where 
k = 1, the slopes are moderate, the color characteristics 

spherical mirror focal rati o = & + d 

lens * 

I At 


| slope 
.too gfreat 



least glass 
to remove 
to maKe) 




K = 4 

gflass too 

These curves may be put on either or both sides of the 
lens provided the value of (t-At) is maintained. 
y4 _ ^ r a V 2 K = a constant between 

4(n-l)a 3 


O and 4 

index of refraction 

of lens material 

Fig. 41. 

are good, and the curve can be put half on one side of the 
plate and half on the other. In the case where k = 1, the 
curve requires the least glass to be removed. 

The Schmidt lens is made of Uviol glass or even fused 
quartz if the camera is to be used for photography in the 



[Chap. II 

ultraviolet spectrum. The Schmidt camera has been made 
to numerical apertures as fast as //O.6. Such a camera is 
much faster than a camera using a lens of corresponding 
aperture because there are fewer glass surfaces to penetrate, 
and the light losses are correspondingly less. 

The Schmidt plate is ground and polished with a special 
ring tool. Each of the glass facets for grinding or pitch 
facets for polishing is mounted on a separate spring as in 

bach of tool 

fa.ce of tool 

Phosphor bronze sectors 

These should be cut from 
the sheet metal in the 
same orientation with 
respect to the rolling of the 
metal in order to have 
their ''springiness" symmetrical. 

For gfrinding, glass or tile 
facets are cemented on 
the spring tips. Tor polishing 
pitch facers are used. 

Fig. 42. 

Fig. 42. This or a similar flexible construction of a tool is 
used since considerable deformability is required of it. 

The curve k = 1 requires the thickness at the center to be 
the same as that at the edge. This is indicated when a 
straightedge laid across the work will touch the center but 
not rock on it. The minimum thickness of the plate at 
y = 0.707 r is determined from the thickness at the edge, 
and the value of At calculated from the equation. The 
intermediary zone is depressed by grinding until this mini- 
mum thickness corresponds with that required as measured 
by a micrometer. 

Figuring may be guided by several testing schemes shown 
at the bottom of Fig. 28. 


Technique of High Vacuum 1 

SOME of the equations from the kinetic theory are im- 
portant in the design, construction, and operation of 
vacuum apparatus. Accordingly, we will begin our treat- 
ment of the technique of high vacuum with a discussion of 
them. The derivations of these equations are omitted, 
since we are interested only in their applications. 

The laws of ideal gases. The laws of ideal gases are 
represented, mathematically, by Eqs. 1 and 2. 

P T = P1 + P2 + • • • Pn< (2) 

Pi represents the total pressure exerted on the walls of a 
vessel containing w\ grams of a gas of molecular weight M h 
when this vessel has a volume V and is maintained at an 
absolute temperature T. If more than one gas is present, 
for example, if the vessel contains Wi grams of one gas of 
molecular weight M h w 2 grams of a second gas of molecular 

1 This chapter is intended primarily to supplement the works on vacuum 
technique listed below: 

Dunyoer, L., Vacuum Practice. New York: D. Van Nostrand and Com- 
pany, 1926. 

Dushman, S., Frank. Inst, J., 211, 689 (1931). 

Dushman, S., High Vacuum. Schenectady: General Electric Company, 

Goetz, A., Physik und Technik des Hochvakuums. Aktges. Braunschweig: 
Friedrich Vieweg und Sohn, 1926. 

Kaye, G. W. C, High Vacua. New York: Longmans, Green and Com- 
pany, 1927. 

Newman, F. H., The Production and Measurement of Low Pressures. New 
York: D. Van Nostrand and Company, 1925. 



weight Mi, and so forth, the partial pressure exerted by each 
gas is given by Eq. 1. 

The total pressure, given by Eq. 2, is the sum of these 
partial pressures. The value of the constant R, the so- 
called universal gas constant, is independent of the molecular 
weight of the gas, but its value does depend on the units in 
which the pressure and volume are expressed. In vacuum 
work the pressure is usually expressed in millimeters of 
mercury 2 and the volume in cubic centimeters, in which case 
R has the value of 62,370. 

Eqs. 1 and 2 are based on the assumptions, first, that the 
molecules are infinitely small and, second, that no inter- 
molecular forces exist. Neither assumption is valid for 
real gases. Nevertheless, the equations describe the be- 
havior of real gases, especially hydrogen and helium, with 
sufficient accuracy for our purposes here. Although the 
equations break down at elevated pressures (pressures 
greater than 1 atmosphere), they become increasingly precise 
if the pressure is reduced. And, at pressures encountered 
in vacuum work, Eqs. 1 and 2 not only apply to the descrip- 
tion of the behavior of gases but describe the behavior of 
many unsaturated vapors as well. 

The mean free path. The mean free path is the average 
distance traversed by molecules between successive inter- 
molecular collisions. The magnitude of this quantity is 
determined by the size of the molecules and is given by the 

x = vib- (3) 

o- represents the molecular diameters and n the number of 
molecules per cubic centimeter. Values of the mean free 

2 P is usually expressed by physicists in millimeters of mercury pressure. 
Other units are the following: 

1 millibar = 0.75 mm 

1 Tor = 1 mm 
1 micron = 10~ 3 mm 



path for nitrogen calculated by Eq. 3, using 3.1 X 10 8 cm 
for the molecular diameters, are given in Table I. 

Mean Free Path of Nitrogen at 0° C. 

Pressure in Millimeters 

Mean Free Path 

of Mercury 


8.5 X 10" 6 cm 


0.0065 cm 

10" 3 

6.5 cm 

10~ 4 

65 cm 

10" 5 

6.5 m 

10" 6 

65 m 

10" 9 

65,000 m 

Viscosity and heat conductivity. The viscosity and heat 
conductivity of a gas, like the mean free path, depend on the 
molecular diameters. As a result, we have the relationship 
between the mean free path and the viscosity 77, 

v = ipv av \, 


and the relationship between the viscosity and the thermal 
conductivity K, 

K = no* (5) 

In these equations p is the gas density in grams per cubic 
centimeter; c v is the heat capacity at constant volume of 
unit mass of the gas; and e is a constant, being 2.5 for mona- 
tomic and 1.9 for diatomic gases. v av . is the average velocity 
of the molecules and is defined by the equation 


X 10 S T 




The relationship between o-, X, rj, and K for various gases is 
illustrated in Table II. 




Properties of Gases 




(<r X 10 8 cm) 

of Viscosity 
(V X 10 6 cm) 

(K X 10 6 cm) 













Kaye, G. W C, and Laby, T. H., Tables of Physical and Chemical Constants 
and Some Mathematical Functions. New York: Longmans, Green and Com- 
pany, 1936. 

Substituting PM/RT for p and Eq. 3 for X in Eq. 4, we 
see that the pressure cancels. In other words, Eq. 4 pre- 
dicts that the viscosity will be the same at reduced pressure 
as it is at ordinary pressures. The experimental verification 
of this prediction by Meyer and Maxwell was a triumph for 
the kinetic theory. 3 They measured the damping of a 
torsion pendulum in a bell jar at pressures varying from 1 
atmosphere to about 10 mm of mercury. The damping 
produced by the viscosity of the air was found to be the 
same at all pressures. 

Eq. 5 predicts that the heat conductivity is also indepen- 
dent of the pressure. This was established experimentally 
by Stefan. 4 

Eqs. 4 and 5 are derived from the assumption that the 
mean free path is small in comparison with the size of the 
apparatus. Table I shows the pressures at which this 
assumption becomes invalid. 

If Meyer and Maxwell had reduced the pressure in their 
bell jar below about 10 _1 mm, they would have observed a 
decrease in the damping effect on the torsion pendulum. 

3 Meyer, O., and Maxwell, James Clerk, Pogg. Ann., 125, 40, 546 (1865), 
143, 14 (1871). 

4 Stefan- J., Akad. Wiss., Ber., 65, 2. 45 (1872). 


Likewise, if Stefan had extended his observations, he would 
have found a decrease in the heat conductivity towards 
10 -1 mm and its complete disappearance below about 
10~ 4 mm. 

Pumping speeds. Consider that a vessel contains a gas 
at pressure P and opens through an aperture to a region 
where a high vacuum is maintained. Further assume that 
this high vacuum is to be maintained at a pressure so much 
lower than P that it is essentially a perfect vacuum. The 
volume of gas escaping through the aperture per unit time, 
dV/dt, measured at pressure P, is given by the formula 

= aJi.S2 X 10 7 -^cm 3 /sec, (7) 

where A is the area of the aperture. The value of dV/dt for 
air (M = 29) at room temperature (T = 300° Kelvin) is 
11,700 cc/sec. cm 2 , or 11.7 liters/sec. cm 2 . It is a note- 
worthy feature of this formula that dV/dt is independent of 
the pressure in the vessel. 

A hypothetical aperture of unit area communicating with 
an essentially perfect vacuum may be regarded as a pump 
with a speed of 11.7 liters/sec. Oil and mercury diffusion 
pumps have two characteristics in common with such an 
aperture. They have pumping speeds of the same order of 
magnitude as the aperture, and their observed pumping 
speeds are roughly constant over a considerable pressure 

The speed of a diffusion pump is, accordingly, expressed as 
the volume of gas passing through the throat of the pump 
measured at the pressure which obtains at the throat. The 
speed factor of a pump is the ratio of its speed per unit area 
of the throat to the value 11.7 liters/sec. A good oil 
diffusion pump has a speed factor of about 0.5 or 0.6. The 
speed factor for mercury diffusion pumps 5 varies from 0.1 
to 0.3. 

? Ho. T. L., Rev. Set. Instruments, 8, 133 (1932). 



mercury pellet 

glass capillary 

The pumping speed of diffusion pumps can be measured 
by means of a leak like the one shown in Fig. 1. Gas at 
atmospheric pressure is allowed to leak into the pumping 
line. The rate at which the gas is introduced is measured 

by the motion of a mercury 
pellet in the calibrated capillary 
tube. At the same time the 
pressure at the throat of the 
pump is determined with a 
vacuum manometer. The rate 
dV/dt at which gas passes 
through the pump is obtained 
by multiplying the volume 
which the mercury pellet sweeps 
through per unit time by the 
ratio of the pressure in the cap- 
illary (that is, the barometric 
pressure) to the pressure which 
obtains at the pump throat. 

Conductance of vacuum 
pumping lines. Ordinarily, a 
pump is connected to an ap- 
paratus by a tube or system of 
tubes which constitute the vac- 
uum pumping line. The meas- 
ured speed of the pump, which 
we will designate S , at one 
end of the vacuum line is greater than the effective pump- 
ing speed, S, at the other end of the line. Naturally, the 
difference between S and S is small if the pumping tubes 
are short and have a large diameter. The difference be- 
tween aS _1 and $o -1 determines the capacity of a vacuum line. 
The capacity is the reciprocal of W, the resistance of the 
vacuum line to the flow of gas. The relationship of the 
quantities S , S, and W is given by the formula 

fine thread 

sewing* needle 

to vacuum 

Fig. 1. 




W, in turn, is defined in terms of tube dimensions by Knud- 
sen's formula, 

W = 1.59X10-^(1 + ^)^ (*) 

where I is the length of the pumping line and d is its diameter, 
both expressed in centimeters. 6 The first term in the par- 
entheses represents the resistances of the line, .. while the 
second term represents the resistance of the two ends of the 
line (or the resistance of a sharp bend in the line). The 
second term is usually insignificant in comparison with the 
first and may be neglected. For example, W. Klose found 
that a straight pumping channel with four right-angle 
bends, one with four T-shaped enlargements, and a curved 
tube of equal diameter all exhibited essentially the same 
pumping speed. 7 

The coefficient of Eq. 9 becomes unity if 29, the molecular 
weight of air, is substituted for M, room temperature of 
300° K. is substituted for T, and 8r 3 is substituted for d 3 , 
where r is the radius of the tube. It is further required that 
V and r be expressed in millimeters and that W be expressed 
in sec. /liter instead of sec. /cm 3 . After making these sub- 
stitutions and neglecting the second term in the parentheses, 
Eq. 9 reduces to 

W' = - — • (10) 

VV r 3 liter V U; 

An an example of the application of Eq. 10, consider a 
pumping tube of 250 mm length and 5 mm radius. This 
gives a value of W equal to 2 sec. /liter. Substituting this 
value in Eq. 8, we see that the pumping speed S can never 
exceed ^ liter /sec, even if a very fast pump is used, for which 
1/So is practically zero. 

Evacuation. The factors determining the rate at which an 
apparatus is evacuated are the volume of the apparatus, V x 

6 Knudsen, M., Ann. d. Physik, 28, 75, 999 (1908). This formula applies 
when d is less than the mean free path. 
7 Klose, W., Phys. Zeits., 31, 503 (1930). 


the effective speed of the system of pumps, S, and the 
limiting pressure which the pumps are capable of attaining, 
P . The method of evaluating the first factor, V, is obvious. 
The value of S may be calculated from the values of S and 
W by Eqs. 8 and 10, or it may be measured by connecting 
the leak and gauge to the apparatus. 

The value of P is not easy to estimate, so it is necessary 
to measure it with a gauge. P does not depend on the 
pumping speed of the pumps on tight systems which are 
outgassed. When the system is leaking or giving off gas, 
P depends on the rate of leaking as well as the speed of 
the pumps. On a tight outgassed system the limiting 
pressure for mercury diffusion pumps equipped with a liquid 
air trap is 10~ 7 mm or less. For oil diffusion pumps without 
traps the limiting pressure varies from 10 -5 to 10 -6 mm, 
although lower values are occasionally reported. The 
vacuum attainable with mechanical pumps is usually 10~ 2 
to 10 -4 mm. The water aspirator is restricted to work at 
pressures above the vapor pressure of water, about 25 mm 
of mercury at room temperature. 

The effect of outgassing on P is illustrated by an experi- 
ment described by Dushman. 8 He found a limiting pressure 
of 0.033 bar for a Gaede rotary pump connected to a vacuum 
gauge when the connecting glass tube was giving off gas. 
When the glass tubing, however, was baked out until its 
surface was free of absorbed moisture and other gases, the 
limiting pressure was reduced to 0.0007 bar. 

The rate at which the pressure is reduced in an apparatus, 
as determined by the pumping speed S, the volume V, and 
the limiting pressure P , is given by the equation 

? = ~ 7 (F - Po) - (11) 

The integration of this equation yields 

{h-h) ==Ilog s (g^). (12) 

8 Dushman, S , Phys. Rev., 5, 225 (1915). 



Eq. 12 is useful, for example, in predicting the time 
(£2 — £1) required for a vacuum system to recover from a 
surge of gas which raises the pressure to the value Pi. In 
this case P 2 represents the working pressure required in the 

If P 2 and Pi are much larger than P , then P may be 
neglected, and Eq. 12 can be simplified to the form 


Roughing pumps. The so-called roughing pumps are 
used to support diffusion pumps because the latter will 

.5 *> 



-.1 ° 


.05 w 


lO^xZ 5 4$6 7&3lO-»x 6" 10-3 

pressure in m m of mercury mm 

Fig. 2. Pumping speeds of mechanical pumps (data supplied by the 
Central Scientific Company, Chicago, Illinois). 

operate efficiently only against a small differential of pressure 
at pressures less than a few tenths of a millimeter of mercury. 
Rotary mechanical pumps are ordinarily used. 9 The pump- 
ing speeds of several rotary pumps at various pressures are 
given in Fig. 2. Other types of pumps, such as the water 

9 The series High-vac, Mega-vac, and Hyper-vac is supplied by the Central 
Scientific Company, Chicago, Illinois. The Leybold vacuum pumps are 
handled in this country by James G. Biddle Company, Philadelphia, Penn- 
sylvania. Extremely fast mechanical pumps are manufactured by the Kinney 
Manufacturing Company, 3541 Washington Street, Boston, Massachusetts. 


aspirator, the Gaede rotary pump, and the Sprengel and 
Toepler pumps, are seldom used now. These pumps are 
adequately described in the literature. 10 

Outgassing of glass and metals. Outgassing removes 
gases adsorbed to the surface of glass and metal. It is 
necessary to outgas exposed glass and metal in order to ob- 
tain the highest degree of vacuum. Prolonged heating of 
glass at 150° to 200°C. in vacuum removes most of the gases 
adsorbed on the surface, while further heating to 300 °C. 
removes the final monomolecular film of water and ad- 
sorbed gases. Gases liberated when the heating is carried 
above this temperature originate from the decomposition 
of the glass. 11 ," 

In practice, lead-glass apparatus is outgassed by heating 
it in an oven or with a soft flame to a maximum temperature 
of 360°C. for a time varying from 10 minutes to an hour or 
more. Lime glass and hard glass are heated to 400° and 
500 °C. respectively. Higher temperatures are to be avoided, 
since the annealing or softening point of soft glass is only 
425°C. and of hard glass 550°C. 

Before a glass apparatus is sealed off from the pumps, the 
seal-off constriction is heated for a minute or two at a tem- 
perature just below the softening point of the glass. 

When metals are strongly heated in a vacuum, they give 
off adsorbed gases as well as absorbed gases and gas arising 
from the decomposition of oxides near the surface. Gases 
under the surface layer of the metal, both dissolved gases 
and those held in chemical combination, are difficult to re- 
move, even at elevated temperatures, unless the metal is 
fused. The metal oxides, with the exception of chromium 
oxide, are readily dissociated in vacuum at elevated tempera- 
tures. Metals which have been fused in vacuum are now 
available commercially. 12 

10 See footnote 1, page 93. 

11 R. G. Sherwood's report on decomposition of glass: Am. Chem. Soc., J., 
40, 1645 (1918); Phys. Rev., 12, 448 (1918). 

12 These metals may be obtained from the Eisler Corporation Newark, New 


Surface gas on tungsten wire is liberated by a temperature 
of 1500 °C. From 70 to 80 per cent of this gas is carbon 
monoxide, and the remainder is hydrogen and carbon diox- 
ide. 13 The volume of surface gas evolved, measured at 
standard conditions, amounts to three or four times the 
volume of the tungsten wire. Sweetser studied the gas 
liberated by copper, nickel, Monel, and copper-coated 
nickel-iron alloy (Dumet). He found that these metals 
rarely gave off a volume of gas greater than the volume of the 
wire. 14 

Marshall and Norton have studied the gases given off 
by tungsten, molybdenum, and graphite. 15 After these ma- 
terials have been outg&osed by prolonged heating in vacuum 
at temperatures above 1800°C, they may be exposed to 
atmospheric pressure, and the gases which they then take up 
are readily removed by subsequent reheating to a moderate 
temperature in vacuum. However, they should not be 
touched with the fingers. j 

Many metals may be heated in hydrogen to remove 
surface contamination. At the same time dissolved gases 
near the surface of the metal are, in part, replaced by the 
hydrogen. This substitution is desirable, since hydrogen 
comes off readily when the metal is subsequently heated 
in vacuum either in a bake-out oven or by high-frequency 

Vapor pressure of waxes. Table III gives the results of 
Zabel's measurements of the relative vapor pressures of 
waxes used in vacuum work. The numbers given there 
represent the results of measurements taken with an ioniza- 
tion gauge. 

The wax compounded from shellac and butyl phthalate 
(see Chapter XIII) should exhibit a low vapor pressure, 
judging from Table III. 

13 Langmuir, I., Amer. Inst. Elect. Engin., Proc, 32, 1921 (1913). 

14 S. P. Sweetser's results are reviewed in Dushman's High Vacuum, 
page 163. 

"Norton, F. J., and Marshall, A. L., Reprint No. 613, General Electric 
Company (1932). 


Vapor Pressure of Material Relative to That of Brass 




Iron coated with rust 


Beeswax and rosin , 


DeKhotinsky (soft to hard) f 
Glyptal lacquer (baked) 

Butyl phthalate 

Stopcock grease 

Ramsay Fett 






15 to 25 




Zabel, R. M., Rev. Sci. Instruments, 4, 233 (1933). 

* Sager, T. P., and Kennedy, R. G., Jr., Physics, 1, 352 (1931). 
t Old formula. A new wax is now supplied by Central Scientific Company 
for which these values may not apply. 

Getters. Ordinarily, in the laboratory, a diffusion pump 
is used to remove the residual gases which roughing pumps 
cannot remove, and the resulting high vacuum is maintained 
by continued pumping. There are, however, other methods 
of removing the residual gases in an apparatus which is 
sealed off at the pressure attainable with a roughing pump. 16 
These methods involve the use of so-called getters, which 
not only remove the residual gases initially, but maintain 
the vacuum against the deteriorating effects of subsequent 

Getters may be grouped into three classes, depending on 
the manner in which they remove residual gases. Some 
depend on the physical adsorption of the residual gases on 
the refrigerated surface of a porous substance like charcoal 
or silica gel; others absorb the gas in the manner that hydro- 
gen is absorbed by palladium black or tantalum; and still 
others combine with the residual gas chemically. 

16 Andrews, M. R., and Bacon, J. S., "Systematic Investigation of the 
Action of Getters in Sealed Tubes," Am. Chem. Soc, J., 53, 1674 (1931). 



The high absorbing capacity of charcoal and silica gel is 
due in part to their large surfaces. The surface of charcoal, 
for example, is estimated to be as great as 2500 square 
meters per gram. Absorbent charcoal to be used for re- 
moving residual gas is itself first outgassed by heating it in 
the vacuum produced by the roughing pumps. It should 
not be heated above the softening temperature of Pyrex, be- 
cause it will lose some of its absorption capacity owing to 
" crystallization" of the charcoal and attendant loss of sur- 
face area. After this outgassing the pumps are turned off 
to isolate the vacuum system, and the charcoal is cooled 
(preferably with liquid air) to develop its absorbing capacity. 
The absorbing power of charcoal for various gases at 0°C. 
and — 185 °C. (liquid air temperature) is given in Table IV. 


Absorption Capacity of Coconut Charcoal: Volume of Gas at Standard 

Conditions of Temperature and Pressure Absorbed by 

Unit Volume of Charcoal 







Carbon dioxide 















Dewar, Sir James, Encyclopaedia Britannica, 16, 751 (1911). 

Of the metal getters, tantalum is of special interest. It 
absorbs hydrogen in large volumes — it may absorb as much 
as 740 times its own volume of gas at temperatures around 
600 °G. This absorbed gas is given off when the metal is 
heated in vacuum at temperatures greater than 800 °C. At 
high temperatures, tantalum is one of the metals most easily 
outgassed. At elevated temperatures the residual gases, 
oxygen and nitrogen, are also removed by chemical combina- 


tion with tantalum. Because of these properties, it is fre- 
quently used for radio-tube anodes. The metals columbium 
and zirconium behave in much the same way as tantalum. 

Tungsten and molybdenum, at temperatures above 
1000 °C, are effective getters. 17 Oxygen is removed by these 
metals by the formation of oxides which are volatile at 
temperatures above 1000 °C. Hydrogen is dissociated by the 
high temperature and condenses as atomic hydrogen on the 
container walls, especially if they are cooled with liquid air. 

The alkali metals react with nitrogen, oxygen, hydrogen, 
and mercury vapor. The absorption of nitrogen, oxygen, 
and hydrogen is especially strong when the alkali metal is 
the cathode of a glow discharge. 

Barium, calcium, and magnesium are extensively used as 
getters, since they combine chemically with all residual 
gases (noble gases excepted). Barium is more active chemi- 
cally than calcium. These metals are introduced by various 
ways into the vacuum tubes in which they are to serve as 
getters. Calcium may be introduced in the form of fresh 
filings. Barium may be introduced in the form of copper- 
or nickel-covered wire. Either metal may be formed directly 
in vacuum by reducing it at elevated temperatures from one 
of its compounds. Usually the introduced metal is vaporized 
and condensed on the walls of the sealed-off vacuum system, 
where it forms a mirror. The getter action of the metal is 
greater in the vapor phase, although the condensed mirror 
film, especially a film of barium, will react chemically with 
residual gases which may subsequently appear in the appa- 

A metal film exhibits, in addition to the chemical action, 
a physical action which may be of considerable significance. 
This physical action, the adsorption of gases, is strong be- 
cause the metal surface is clean. Dushman gives an ele- 
mentary calculation illustrating this action. 18 A spherical 

17 Langmuir, I., Am. Chem. Soc, J., 37, 1139 (1915); Indust. and Engirt. 
Chem., 1, 348 (1915). 

18 Dushman, S., Frank. Inst, J., 211, 737 (1931). 


bulb 5 cm in radius containing residual gas at a pressure of 
about to nun of mercury will be completely evacuated when 
sufficient gas is adsorbed on the inside surface of the bulb 
or on a clean metal film to form a monomolecular layer. 

Water and many vapors may be effectively removed by a 
trap cooled in liquid air. The density of water vapor in a 
gas, after it is passed through a liquid air trap, is 10~ 23 
mg/liter. The relative effectiveness of some of the more 
commonly used drying agents is shown in Table V. 19 Of 

Drying Agents 

Drying Agent 

Mg of Water per Liter of Gas Dried 
at 25°C. 

Trap at liquid air temperature 

P 2 5 

1.6 X 10" 23 
< 2 X 10" 6 

Mg(C10 4 ) 2 

Mg(C10 4 ) 2 -3H 2 

< 5 X 10" 4 

< 2 X 10~ 3 

H 2 S0 4 

3 X 10~ 3 

95 per cent H 2 S0 4 

CaCl 2 (gran.) 

3 X lCT 1 

1.4 to 2.5 X 10" 1 

National Research Council, International Critical Tables, Vol. Ill, page 385. 
New York: McGraw-Hill Book Company, 1928. 

these, phosphorus pentoxide is the one most frequently used 
in vacuum work. It should be fused to reduce its vapor 
pressure and to prevent it from flying about when the 
system is evacuated. 

Static and kinetic vacuum systems. Most of the vacuum 
systems used in physical research fall into two general classes. 
In the first class we have those systems which are required 
to be thoroughly outgassed and entirely free from leaks in 
order to obtain a high degree of vacuum. We will call 
systems of this type static vacuum systems, in contrast to 

19 A drying agent which has the advantage of being solid when it is saturated 
as well as when it is "dry" is magnesium perchlorate. This chemical is manu- 
factured by the Arthur H. Thomas Company, Philadelphia, Pennsylvania. 

two-stage mercury 
diffusion pump with 

_j 'poJ a L cub 'l_heater 

SceJe of Feet 

6 inches O 

Fig. 3. 



systems in which outgassing from glass on metal parts or in 
which even small leaks may be tolerated, owing to the use 
of extremely fast pumps. We will designate systems of the 
latter type as kinetic vacuum systems. 

Fig. 3 illustrates a typical static vacuum system. It 
represents an X-ray tube being evacuated with a mercury 
diffusion pump of moderate speed. Pressures as low as 
10~ 9 mm (or even 10~ 10 mm) are obtained in some static 
vacuum systems. Such extremely high vacuum is required 
for investigating the photoelectric effect, thermionic emis- 
sion, and other physical phenomena for which the slightest 
contamination of a surface is to be avoided. Static vacuum 
systems are not treated extensively here. The reader who 
is especially interested in them is referred to the literature. 

Kinetic vacuum systems are characterized by a limiting 
pressure of 10~~ 5 to 10 -6 mm obtained by the use of extremely 
fast pumps. These pumps, as well as the apparatus which 
they exhaust, are usually made in the machine shop from 
ordinary brass and steel. The metal is not outgassed as in 
static vacuum systems. 

Kinetic vacuum systems are inferior to static systems, 
where surface contamination must be scrupulously avoided. 
They are, however, satisfactory for applications where the 
function of the vacuum is to allow the unhindered motion of 
molecular rays, electrons, ions, and light quanta. For ex- 
ample, kinetic vacuum systems have been applied with 
success to the vacuum evaporation process for metalizing 
large telescope mirrors, to the maintenance of vacuum in 
high- voltage X-ray tubes, metal rectifier tubes, and oscillator 
tubes, and to the evacuation of spectrographs. 

Fig. 4 shows a kinetic vacuum system for the metalization 
of glass mirrors. There are two obstacles in the way of 
getting a high vacuum in such a system. First, outgassing 
by heating is precluded on account of the use of wax seals 
and on account of the fact that the system may contain thick 
glass mirrors which cannot be safely heated. Second, 
there is more chance of small leaks appearing than in a 




cenirit plank of table 
removable to allow 
access "to pomp* 

gloss stop- 

to mech anic al 

pomps wsjJi 

6" oil-diffusion 
pump with 
Hot spiral 


2 feet 

Fig. 4. 



static vacuum system, since the system shown in Fig. 4 
must be repeatedly opened. The recent development of fast 
oil diffusion pumps, which give the degree of vacuum required 
in spite of these obstacles, has been mainly responsible for the 
modern extensive use of this type of flexible vacuum system. 

Diffusion pumps. Diffusion pumps will operate only if 
the pressure is less than a few tenths of a millimeter of mer- 
cury, and they operate best with a "backing pressure" of a 
few hundredths of a millimeter of mercury. The necessary 
"backing pressure" is obtained by mechanical pumps. The 
operation of a mercury diffusion pump is illustrated in Fig. 5. 
The pump shown here illustrates Langmuir's practical adap- 
tation of Gaede's discovery of the principle of diffusion 
pumping. 20 The following explanation of its action applies 
as well to the action of oil diffusion pumps. 

A stream of mercury vapor is obtained by heating liquid 
mercury in boiler B to a temperature of about 110°C. 
The vapor stream which 
effuses from the attached 
chimney is indicated by 
arrows. This stream 
forms a partition be- 
tween chamber N and 
chamber M. The vapor 
finally condenses on the 
water-cooled walls of 
chamber N and returns 
under the influence of 
gravity to the boiler as a 
liquid. Gas molecules in 
chamber N which diffuse 


air press utre 
0.3 mm mercury 
or less for 
pumping action 
to start 

water cooling 




Fig. 5. Diagrammatic sketch of 
Langmuir's diffusion pump. 

into the vapor partition have a small chance of penetrating 
it and entering chamber M. Rather, it is more probable 
that they will be carried by the stream back into chamber N. 
However, gas molecules in M which diffuse into the vapor 

20 Langmuir, I., Phys. Rev., 8, 48 (1916). 
Gaede, W., Ann. d. Physik, 46, 357 (1915). 




Fig. 6. 



partition are carried along by molecular bombardment into 
N, where they are removed by the mechanical pump. 

The pressure in N must exceed 
that in M by a factor of the order 
of 100 if the rate of diffusion is 
to be the same in both direc- 
tions across the vapor partition. 
Where N is evacuated by an aux- 
iliary diffusion pump instead of 
the mechanical pump, pressures 
of 10 -7 mm of mercury or lower 
can be obtained in a tight glass 
apparatus connected to M (pro- 
vided mercury vapor is removed with a liquid air trap). 

Mercury pumps have been studied by many investi- 
gators. 21 Figs. 6 to 12 are 
representative of the de- 
signs which have evolved 
as a result of these studies. 
We will not discuss these 
pumps in detail, as we are 
mainly interested in this 
chapter in kinetic vacuum 
systems and oil diffu- 
sion pumps. With oil 
pumps it is not uncommon 
to have pumping speeds of 
some tens or hundreds of 
liters per second, whereas 
with mercury diffusion pumps the speeds are ordinarily only 
a fraction of a liter per second up to a few liters per second. 

2i Crawford, W. W., Phys. Rev., 10, 558 (1917). 
Klumb, H., Zeits.f. techn. Physik, 17, 201 (1936). 
Molthan, W., Zeits. f. techn. Physik, 7, 377, 452 (1926). 
Stintzing, H., Zeits.f. techn. Physik, 8, 369 (1922). 

See the references to vacuum technique given in footnote 1, page 93, and 
other references cited herein. See also catalogues of E. Leybold Nachfolger. 
Gaede, W., Zeits. f. techn. Physik, 4, 337 (1923). 
Ho, T. L., Rev. Sci. Instruments, 3, 133 (1932); Physics, 2, 386 (1932). 

Fig. 7. Crawford's diffusion pump. 



The use of oils as diffusion pump liquids. There have 
been many attempts to find a substitute for mercury as a 
pumping medium, for the use of mercury has one consider- 
able disadvantage, namely, its vapor pressure is so high that 
traps are required to prevent it from diffusing into the vac- 
uum system and destroying the vacuum. These traps, 
having a high resistance to the flow of gas, choke the pump. 
The only widely used substi- 
tutes for mercury are oils. The 
oils used for this purpose are 
either especially refined petro- 
leum oils of the naphthene type 
as developed by C. R. Burch, 22 
or they are organic compounds 
such as butyl phthalate as de- 
veloped by Hickman and San- 
ford 23 of the Eastman Kodak 
Laboratories. Recently, Hick- 
man has recommended a new 
synthetic organic oil called 
Octoil, which is claimed to be 
superior to butyl phthalate. 24 
Oils of the type developed by 
Burch are manufactured under 
Metropolitan Vickers' patents under the trade name of 
Apiezon oil. 25 Similar oils are now available in this country 
which yield pressures below 10 -6 mm of mercury. 26 

Oil pumps have the advantage over mercury pumps that 
they do not require traps except in certain applications. 

22 Burch, C. R., Nature, 122, 729 (1928); Roy. Soc, Proc, 123, 271 (1929). 

23 Hickman, K. C. D., and Sanford, C. R., Rev. Sci. Instruments, 1, 140 

2 4 Hickman, K. C. D., Frank. Inst., J., 221, 215, 383 (1936). 

25 This oil may be obtained from the James G. Biddle Company, Philadel- 
phia, Pennsylvania. 

26 Relative to pump oils see the following: 

von Brandenstein, Maruscha, and Klumb, H., Phys. Zeits., 33, 88 (1932). 
Klumb, H., and Glimm, H. O., Phys. Zeits., 34, 64 (1933). 
These oils may be obtained from Litton Laboratories, Redwood City, 
California, and the Central Scientific Company, Chicago, Illinois. 

Fig. 8. Down-jet diffusion pump. 



Another advantage is that oil pumps may be fabricated 
either from steel or from brass and copper, whereas metal 
mercury pumps must be constructed of steel with welded 
joints. Brass and copper pumps can be assembled with soft 

solder, except for the boiler 
and chimney, where it is ad- 
visable to use silver solder. 
Aside from the questions of 
traps and construction, the 
contrast between oil and 
mercury pumps is less dis- 
tinct. Oil pumps without 
traps do not give quite as 
low a limiting pressure as 
trapped mercury pumps, al- 
though their speed may be 
many times greater. If 
traps are used, there is 
probably little difference 
between the limiting pres- 
sures attainable. Oil 
pumps have the advan- 
tage that a baked-out total 
obstruction charcoal tube 
at room temperature is as 
effective as a liquid air 
trap. However, the use of 
a total obstruction charcoal 
trap sacrifices the higher 
pumping speed of the oil pump. 

It is not advisable to use a single oil pump. One should 
use at least two oil pumps in series. The second pump serves 
to keep the oil in the first purified. The limiting pressure 
is about tenfold lower when a second pump is used. Be- 
cause mercury pumps will operate against a slightly higher 
back pressure than oil pumps, there are many cases in which 
a single mercury diffusion pump is adequate. 

electric heater 

Fig. 9. Ruggles' and Kurth's two- 
stage mercury diffusion pump. 




Oil diffusion pumps. Oil diffusion pumps are like mer- 
cury diffusion pumps in several respects. They have the 
same functional elements — a boiler to vaporize the oil and a 
chimney for conducting the vapor to the jet. The two 
types of pumps are also similar in the way in which they 
function. The oil vapor is projected from the jet across the 
throat of the pump and condenses 
on the cooled walls which form 
the outer boundary of the throat; 
and the condensed oil drains 
from the condensing surface back 
into the boiler by gravity. The 
vapor jet may be arranged in 
several ways: It may be directed 
upward as in the up-jet mercury 
pump shown in Fig. 5, it may be 
directed downward as in the um- 
brella down-jet mercury pump 
shown in Fig. 6, or it may pro- 
ject laterally as shown in Fig. 7. 

Although oil and mercury 
diffusion pumps have the same 
functional elements, they differ 
in the details of construction. 
The construction of oil diffusion 
pumps can be carried out in an 
ordinary machine shop. The important considerations for 
proper construction are outlined below: 

1. The oil is decomposed slightly at the working tempera- 
tures of the boiler. This decomposition is accelerated by the 
higher temperature necessary when the cross section of the 
boiler is not large enough to afford an adequate surface from 
which to create vapor, or when the chimney and jet are not 
ample to deliver the required amount of vapor without an 
excessively high pressure drop. 

2. Since oil has a low latent heat, the pump should be 
designed so that the heat required to maintain the working 

Fig. 10. 



Pl&nof Jets 

temperature of the chimney and jet is supplied by conduc- 
tion from the heater rather than by condensation of oil 
vapor. Naturally, copper is the best ma- 
V>«* terial for constructing the chimney on 
|^« account of its large heat conductivity. 

3. The decomposition of the oil is cata- 
lyzed by copper and brass and not by 
nickel. Accordingly, all parts of the 
pump exposed to the hot oil should be 
nickel-plated. 27 

4. The amount of oil decomposed in a 
given time is proportional to the amount 
of oil present in the boiler. It is, there- 
fore, advisable to have only a shallow 
layer of oil in 
the boiler. 

5. At least 
two single -jet 
pumps in series 
should be used. 
pumps are not 

recommended because of the diffi- 
culty of regulating the flow of va- 
por to the various jets and of sup- 
plying the necessary amount of 
vapor required by them without 
an excessive boiler temperature. 

6. Throat clearances narrower 
than | inch are practical only for 
up-jet pumps. Condensed oil will 
bridge gaps of this narrowness in 
pumps of the down-jet type. 

7. Backward evaporation of the oil into the pumping line 
should be restrained by the use of baffles. 

27 Privately communicated: Charles V. Litton, Engineering Laboratories, 
Redwood City, California. 

Detail of 
single jet 

Fig. 11. 




8. Cold oil is a better solvent for many gases and vapors 
than hot oil. Accordingly, the condensed oil should be 
returned to the boiler at the maximum temperature possible. 
Otherwise, a certain amount of the exhaust gases and vapors 
dissolve in the condensed oil 
and contaminate it. 

9. The use of electric heat 
for the boiler is advisable, 
since it is subject to more 
delicate control than gas 
heat. A Calrod heater unit, 
such as used in electric stoves, 
can be re-coiled into a helix 
of 2 inches in outside diameter 
or as a flat spiral of smaller 

Figs. 13 to 18 illustrate 
several oil pumps which are 
currently popular. 28 The 
pump shown in Fig. 13, de- 
signed by Sloan, Thornton, 
and Jenkins, satisfies the re- 
quirements for good design 
outlined above and at the same 
time combines these features 
together with simplicity of 
construction. The following 
description of this pump is a quotation from a paper of 
Sloan, Thornton, and Jenkins. 29 

The Apiezon oil diffusion pump was originally developed by 
the Metropolitan Vickers Company in England for this very pur- 

28 References to pumps having interesting construction but not represented 
here include the following: 

Copley, M. J., Simpson, O. C, Tenney, H. M., and Phipps, T. E., Rev. Sci. 
Instruments, 6, 265, 361 (1935). 

Esterman, I., and Byck, H. T., Rev. Sci. Instruments, 3, 482 (1932). 
Ho, T. L., Rev. Sci. Instruments, 3, 133 (1932); Physics, 2, 386 (1932). 

29 Sloan, D H., Thornton, R. L., and Jenkins, F. A., Rev. Sci. Instruments, 6, 
80 (1935). 

Fig. 13. 



pose of continuously exhausting radio tubes. The oil is sold 
commercially in this country. 

Fig. 13 is typical of the simplified designs which have been 
widely adopted in this country. The outer shell 2" in diameter 
consists of a water-jacketed brass cylinder with a copper plate 
silver-soldered into its bottom. In the cavity beneath the bottom 
plate is placed an electric heater which boils the Apiezon "B" oil at 
less than 200°C in the chamber above. The oil vapor rises 

through the copper chimney 
and is deflected downward 
by a spun copper umbrella. 
The ye-" clearance between 
the edge of the umbrella and 
the condensing wall is not 
critical, although an opti- 
mum exists for any specified 
set of pressures. Around 
the chimney is a glass heat 
shield, and a metal baffle 
plate to retard the rise of oil 
vapor from the roof of the 
boiler, but these can be 
omitted without serious 
consequences. The two 
baffles above the umbrella 
prevent the escape of oil 
vapor directly into the re- 
gion being evacuated. The 
convenient baffle system shown here reduces the speed of the pump 
to less than half, so that its overall speed is only thirty liters 
per second. This is more than sufficient for these oscillator 
tubes, since the connecting system reduces the speed to less than 
ten liters per second. A pressure in the oscillators of 10~ 5 mm is 

Incidentally, the same general design is also well suited to larger 
pumps of 4" and 6" diameter, for use with larger tubes. The 
speed of an oil pump can be greatly increased by enlarging the 
diameter of the overhead region which contains the baffles neces- 
sary to guard against escaping oil vapor. 

butyl phthalate 
or other 
pumping oils 

Fig. 14. 




^ tubing 

A 2-inch pump of such construction will have a pumping 
speed of about 30 liters/sec, or a speed factor slightly greater 
than 50 per cent. 

If such a high speed is not needed, an up-jet pump may 
serve. Fig. 14 shows Hickman and Sanford's all-glass de- 
sign of an up-jet pump. 

Fig. 15 shows an all-metal 
up-jet pump designed by Ed- 
win McMillan. 30 With the 
boiler temperature adjusted to 
give maximum pumping speed, 
this pump will work at a rate 
of 4 liters/sec. against a back- 
ing pressure of \ mm of mer- 
cury. If the boiler tempera- 
ture is too high, the action 
of the pump will be erratic, 
since returning condensed oil 
interferes with the vapor jet. 

A design combining glass 
and metal construction, devel- 
oped by Joseph E. Hender- 
son, 31 is shown in Fig. 16. He 
reports this pump to be capa- 
ble of working against a back- 
ing pressure of a few tenths of 
a millimeter pressure in con- 
trast to the pressure of about 


mm required for oil pumps 

with a throat opening of \ inch 
or more. Pressures as low as 
10 -8 mm of mercury were obtained 
operated with a charcoal trap. 

Fig. 15. McMillan up-jet pump. 
Hole in upper block (indicated by 
arrow) \ inch in diameter. Outside 
diameter of jet A inch. Jet clear- 
ance A inch. The necessary 
baffles above the jet are not shown. 

with it when it was 

30 Privately communicated. 

31 Henderson, Joseph E., Rev. Sci. Instruments, 6, 66 (1935). 





A pump designed by Zabel with a novel oil heater added by 
James A. Bearden 32 is shown in Fig. 17. The advantage of 
a pump of this design is that it quickly starts working after 
the heater is turned on. 

More recently, K. C. D. Hickman and others have experi- 
mented with pumps in which the oil is continually purified. 33 

Pumps of this type are particularly 
suitable for work with gases and 
vapors which dissolve in the oil or 
decompose it. Fig. 18 shows a pump 
which incorporates some of the re- 
sults of Hickman's investigations. 

Mercury traps. Mercury vapor 
diffuses from a mercury diffusion 
pump into the exhausted vessel un- 
less it is removed in a trap by con- 
densation on a cold surface. Besides 
the inconvenience and expensive ne- 
cessity of requiring a refrigerant, the 
use of traps has the more serious 
result of choking the pump. This 
is especially true for big mercury 
pumps of high speed. For example, 
a mercury pump with a speed of sev- 
eral hundred liters per second at its 
throat may have an effective speed 
beyond the trap of only several tens 
of liters per second. 

The common trap designs for con- 
densing mercury and water vapors 
are illustrated in Fig. 19. Type A, the simplest, is fre- 
quently used for trapping the vapors from a McLeod 
gauge. It is also useful in conjunction with an ioniza- 
tion or Pirani gauge for hunting leaks. Type B, the most 

Fig. 16. 

32 Bearden, J. A., Rev. Sci. Instruments, 6, 276 (1935). 
Zabel, R. M., Rev. Set. Instruments, 6, 54 (1935). 

33 See footnote 24, page 113. 



common type, may be conveniently constructed from 
metal and a simple glass tube as shown at B' ', or it may be 


Beard en type 
heater (helix 
of Ni chrome 
wire immersec 
in oil) 

Fig. 17. 

constructed as shown at B" with a separator or baffle to 

cause the gas to circulate against the cold walls of the glass 

tube. Both types A and B 

are immersed in the refrigerant 

liquid. Types C, C, and C" 

contain their own refrigerant, 

but because of inferior heat 

insulation these traps are less 

economical to keep cold. 

As refrigerant liquids for 
trapping mercury and water 
vapor, either liquid air or dry 
ice in acetone may be used. 
The temperature of the for- 
mer varies from — 190°C. to 
— 183 °C, depending on the ex- 
tent to which the nitrogen has 
been boiled out of the liquid 
air, leaving liquid oxygen. Fig. 18. 


It- reservoir for removal 
of volatile components of 
oil formed in the boiler 



The temperature of dry ice-acetone mixture is about — 78 °C. 
At the temperature of liquid air the vapor pressure of mer- 
cury is 1.7 X 10" 27 mm, while at -78°C. it is 3.2 X 10" 9 mm. 
For trapping water, liquid air temperatures are sufficiently 
low. However, since the vapor pressure of ice is about 
10 -3 mm at —78°, the dry ice-acetone mixture is not suffi- 
ciently cold to trap water vapor effectively. Accordingly, 
when this refrigerant is used for mercury, it is necessary at 

Fig. 19. 

the same time to expose anhydrous phosphorus pentoxide 
in the vacuum in order to remove the water vapor. 

The vapor pressure of the vacuum pump oils used in 
roughing pumps, according to Dushman, is 10~ 3 to 10 -4 mm 
at ordinary temperatures, i of this value at 0°C, and 
negligibly small at the temperature of dry ice or liquid air. 

Carbon dioxide is adequately trapped by traps cooled by 
liquid air, since its vapor pressure, at liquid air temperature, 
varies from 10~ 6 mm to 10~ 7 mm. Carbon monoxide, 
methane, ethane, and ethylene, having considerably higher 
vapor pressures, are not effectively trapped even by a 
liquid air trap. 


Virtual leaks. Gases will condense when their partial 
pressure is above the vapor pressure corresponding to the 
trap temperature. (However, they will re-evaporate later 
when the pumps reduce the pressure to a sufficiently low 
value.) This condensation may give rise to a virtual leak 
if the trap is cooled too soon after the evacuation of a system 
is started. We use the term virtual leak because the system 
appears to have a leak, when it is, in fact, quite tight. As 
an example, consider a system with traps cooled with a dry 
ice-acetone mixture but with phosphorus pentoxide omitted. 
Some of the water vapor originally in the system, both in the 
air and from the walls where it is held adsorbed, will be con- 
densed in the trap. As the evacuation of the system pro- 
ceeds, the pressure will approach a limit of 10~ 3 mm, this 
being the pressure of the water vapor in the trap, and the 
system will exhibit all the "symptoms" of a leak. The 
same effect is encountered if liquid air is put on the system 
too soon. Some of the water vapor will condense on the 
upper regions of the trap walls, and as the liquid air level 
around the trap falls, owing to evaporation, the temperature 
of the water condensed as ice will rise until it begins to 
sublime, producing a virtual leak. On the one hand, these 
ice crystals are too cold to evaporate rapidly and be evac- 
uated by the system (or colder regions of the trap), while, 
on the other hand, they are warm enough to degrade the 
vacuum. Likewise, gases like ethylene may condense in a 
trap cooled by liquid air and degrade the vacuum. 

To avoid virtual leaks, the proper procedure is to keep the 
traps warm until a vacuum is obtained at which mercury 
begins to diffuse into the evacuated apparatus, that is, until 
a pressure of about 10 -2 mm is obtained. Then the tip of 
the trap is cooled until the vacuum reaches its limit, P , 
and finally the trap is immersed in the liquid air to the 
full depth. 

"Oil" traps. The vapor pressures of vacuum-pumping 
oils, such as Apiezon "B" oil, are very low, but gases pro- 
duced by thermal decomposition of the oil may give rise 


to some deterioration of the vacuum and necessitate the use 
of traps. For example, when Bearden evacuated an X-ray 
tube with the diffusion pump shown in Fig. 17, he found that 
a carbon deposit formed on the target of the tube. 34 He 
found, also, that the filaments of the tube deteriorated at an 
excessive rate. However, the use of a refrigerated trap 
greatly reduced these effects. The trap he used was cooled 
with dry ice in alcohol. 

The trap shown in Fig. 20 was designed by Hickman for 
diffusion pumps which use Octoil. 35 According to him, it is 
.sufficient to cool the trap with running water. Electric re- 
frigerator units are sometimes 
used to trap vapors from oil 
pumps. These are, naturally, 
justified only in large and 
permanent installations. 

In ordinary experimental 
work, charcoal traps are satis- 
Fi 20 factory for use with oil diffu- 

sion pumps. Several charcoal 
trap designs are shown in Fig. 21. Of these, the total ob- 
struction trap, A, is the most effective, although it has the 
highest resistance, W, for the gases passing through it. 
Becker and Jay cox suggested a trap of type A. They found 
that a charcoal trap removed oil and condensable vapors to 
such a degree that an ionization gauge indicated a "pressure' ' 
as low as 10 -8 mm of mercury. 36 This has been confirmed 
by Joseph E. Henderson. 37 

When charcoal traps become charged with oil and vapors, 
it is necessary to bake them out. Becker and Jaycox ob- 
served that condensed pump oils are decomposed by baking 
them in contact with charcoal, and that the decomposition 
products are gases. 

34 Bearden, J. A., Rev. Sci. Instruments, 6, 276 (1935). 

35 See footnote 24, page 113. 

36 Becker, J. A., and Jaycox, E. K., Rev. Sci. Instruments, 2, 773 (1931). 

37 Henderson, Joseph E., Rev. Sci. Instruments, 6, 66 (1935). 





heaters to 
bake out 
& charcoal 

Fig. 21. 

Construction of kinetic vacuum systems. Glass was 
formerly used extensively for the construction of vacuum 
apparatus, but now metal has replaced it for many uses. 
Glass as a construction material is characterized by its 
transparency, high electrical insulating quality, and by the 
fact that it is easily cleaned and may be baked out and sealed 
off to give a more or less permanent vacuum. Also, aux- 
iliary parts can be welded to an apparatus without the use 
of any gaskets or sealing wax. These welds are easily 
tested for leaks with a spark. 

Unfortunately, large and complicated apparatus is diffi- 
cult to construct from glass. On the other hand, large 
vacuum systems made of metal are not fragile, and repairs 
and alterations on them can be easily made in the machine 

The metal most frequently used is yellow brass. A 
vacuum-tight apparatus can be made from plates and cylin- 


ders of this metal, screwed together and "painted" on the 
outside with beeswax and rosin mixture; or the plates, 
cylinders, and so forth, may be fitted together with rubber or 
lead fuse-wire gaskets. The brass parts may also be soft- 
soldered or silver-soldered, depending on the temperature 
resistance and strength required. 

Steel apparatus may be soft-soldered, silver-soldered, 
brazed, or welded. Electric welding is quite satisfactory 
for vacuum work if it is done in two or three "passes" with 
shielded electrodes. It is generally less subject to leaks than 
gas welding, and it does not warp the work as much. Steel 
vacuum tanks, especially if they are rusty, are sometimes 
coated on the inside with Apiezon wax "W" to stop leaks 
as well as to offer a surface which does not give off gas. 

Since metal vacuum walls outgas more than glass, small 
leaks are more difficult to find. It is a common procedure 
to coat the outside of metal apparatus with lacquer, which 
seals small leaks and at the same time gives a workmanlike 
appearance to the apparatus. Glyptal is heat resistant. 
For example, it may even be used for coating the outside 
surfaces of diffusion-pump boilers. 

Many things are exposed in kinetic vacuum systems which 
one would not expose in static vacuum systems. Chief 
among them are rubber (especially as used for gaskets), 
waxed packing, beeswax and rosin mixture, Apiezon wax, 
and ordinary machined metal parts which are not outgassed. 

Wood, paints and varnishes, porous cements, and rust 
should not be exposed even in a kinetic vacuum system. 

Rubber hose may be used for connections, and with a 
pinch clamp it serves as a venting device. Rubber should 
not be exposed to high vacuum if pressures of the order of 
10 -6 or less are desired. 

Joints. Two tubes of glass or metal may be butt-joined 
by slipping a wide rubber band over them. The rubber 
surface, including the junctions of the rubber to the tubes, is 
painted with several coats of shellac as shown in Fig. 22. 
This type of joint is easily disconnected. For small tubes, 



a short length of rubber hose makes a convenient connection. 
Rubber tape or strips of raw rubber may also be used. Inas- 
much as rubber is somewhat permeable to some gases and 
gives off hydrogen sulphide and other vapors in vacuum, the 
connected tubes should always 
fit together neatly to decrease 
the area of rubber exposed. 
The joint may be first wrapped 
with sheet aluminum and then 
with rubber. 38 This procedure 
decreases the area of rubber 
exposed. If any considerable area of rubber is exposed, it is 
advisable to boil it in a 15 per cent caustic solution (potas- 
sium hydroxide or sodium hydroxide) to dissolve free sulphur 
and remove talc from its surface. It is then washed with 
water and dried, either with alcohol or by a vacuum pump. 
If rubber tubing becomes porous and checked with age, 
it should be painted on the outside with castor oil. 


Fig. 22. 



Fig. 23. 

Two metal tubes may be joined with flanges which are 
sealed with a tongue and groove joint fitted with a rubber 
gasket as shown in Fig. 23. This construction is recom- 

38 The Central Scientific Company supplies a raw rubber tape for this. 
They recommend the use of a piece of thin aluminum sheeting with it. 




mended where mechanical strength is desired and also where 
the joint must withstand moderate internal pressure. The 
tongue should have the same thickness as the groove to 
within a few thousandths of an inch, so that the rubber 
gasket will not extrude as the pressure for fitting the joint 
is applied. The gasket is cut from a sheet of packing with a 
cutter like the one shown. The rubber gasket is used dry, 
and if the tongue and groove have bright smooth surfaces, 
the joint is sure to be free from leaks. Furthermore, the 
joint exposes very little rubber surface to the vacuum system. 

In another type of joint, 
shown in Fig. 24, a lead fuse 
wire can be used as a gasket in- 
stead of rubber. The gasket in 
this case is a loop of 20-ampere 
fuse wire, butt-welded by 
means of the heat from a match 
and a little soldering flux. The 
circumference of this loop is 
made slightly shorter than re- 
quired and is stretched into 
the groove to make a snug fit. 
The pressure applied in the 
flange flows the lead into inti- 
mate contact with the two elements of the joint. Lead- wire 
joints can be used on systems to operate at elevated tempera- 
tures, since they will hold to higher temperatures than tongue 
and groove joints sealed with rubber. A lead gasket of this 
type is used on the 40-inch bell jar for aluminizing astro- 
nomical mirrors as shown in Fig. 13 of Chapter IV. This 
particular joint has been made more than a hundred times, 
and it has been consistently vacuum-tight. Aluminum 
wire holds to even higher temperatures. 

Seals. It is frequently necessary to make a vacuum-tight 
seal between a glass bell jar and a metal base plate. For- 
merly, stopcock grease was used, applied to the foot of the bell 
jar. This type of seal was not always tight, and the grease 


lead fuse 
wire pressed 
to rnowke 
tight seal 

Fig. 24. 



Fig. 25. 

frequently entered the apparatus and contaminated exposed 
surfaces. A better procedure is to use wax instead of stop- 
cock grease. The bell jar is set on the base plate, both the 
foot of the bell jar and the base plate being clean and dry. 
Beeswax and rosin mixture, 
smoking hot, is then applied 
with a medicine dropper to 
the outer edge of the bell-jar 
flange to effect the seal, as il- 
lustrated by Fig. 25. The bell 
jar can be removed from the 
base plate in the following 
manner: After scraping away 
the wax with a putty knife, 
loosen the jar by striking a 
sharp blow at the top with the 
palm of the hand or by driv- 
ing a razor blade gently under 
the edge of the jar. If a metal bell jar is used, a recess may 
be provided so that the seal can be cracked by prying with a 
screw driver after as much of the wax as possible has been 
scraped away. 

Windows may be sealed over observation ports in a similar 
manner. The wax is applied with the medicine dropper, and 
the seal is effected without sensibly heat- 
ing either the port or the window. 

Windows may be sealed with hard wax. 
It is necessary to heat both the port and 
the window to temperatures above 100°C. 
when hard waxes such as Apiezon "W," 
Picein, shellac, or DeKhotinsky wax are 
used. First the window and port are care- 
fully cleaned, and then the window is 
clamped in the desired position. After being heated to the 
required temperature, the wax is applied to the outside edge 
of the window, from where it will be drawn between the 
window and the port by capillary force. The wax drawn 

Fig. 26. 




Fig. 27. 

under the window forms a thin bonding layer of large area, 
which exposes a minimum surface of wax to the vacuum. 
(See Fig. 26.) 

Fig. 27 shows the procedure for sealing two glass tubes 
together with Picein wax to form a butt joint or telescope 

joint. The procedure here is to 
wrap a soft strip of Picein around 
the warmed glass tubes. This strip 
is molded from a stick of wax after 
it is thoroughly softened. The stick 
of wax is softened by alternately heating it in a Bunsen flame 
until its surface is liquid and withdrawing it to cool until its 
surface solidifies. When the strip is ready and while it 
is still soft it is wrapped around the warmed joint and molded 
as shown in Fig. 27. The wax will not stick to the ringers if 
they are damp. After the glass and wax are cool, a flame is 
applied to fuse the wax superficially and insure tangential 
contact to the tubing. 

Electrodes. In the chapter on glass blowing, we dis- 
cussed the construction details for leading electrical con- 

ductors into glass apparatus. 
In a kinetic vacuum system, 
electrodes are usually 
fastened through holes in 
a metal wall. Construction 
details are shown in Fig. 28 
for high-current conductors 
and in Fig. 29 for high- 
potential conductors. The 
high-current conductor or 
electrode consists of a 
brass screw bolted into the 
vacuum wall, the head and 

pumping channel 

overhang of 
washer to pre- 
vent 'shorting^ 
by condensed 
metallic va- 

vacuum wall 
^% heavy copper lead 
Soldered into slot 

Fig. 28. 

body of the screw being insulated from the metal vacuum 
wall with mica. After the insulation has been tested with 
a lamp, the whole assembly is made vacuum-tight by coating 
the screwhead, insulation, and the local area of the outside 



surface of the vacuum wall with beeswax and rosin mixture 
or with glyptal lacquer. Beeswax and rosin mixture is used 
if the operation temperature is about room temperature. 
Glyptal, after baking to polymerize it, is used for operation 
temperatures up to about 100°C. 

The electrode just described does not have high insulating 
qualities. Where better insulation is needed, a capillary 
glass tube is used in either of the ways shown in Fig. 29. 
In either case the electrode is easily removable for cleaning 
off condensed metallic vapors, for replacement of the glass, 

screw plug 
•^* >*Py**ex capillary 


electrode (S^ 

Fig. 29. 

and so forth. If the conductor wire is tungsten, it may be 
sealed directly to the Pyrex capillary. The capillary, with 
its central conductor, is first sealed in a bored machine screw 
with pure shellac or DeKhotinsky wax. This is then screwed 
through the walls of the vacuum system and made tight with 
beeswax and rosin mixture applied on the outside. 

Valves. Valves are used on the low-vacuum side of diffu- 
sion pumps to prevent oil in the mechanical pumps from 
flowing into the other parts of the apparatus. Between the 
diffusion pumps and the apparatus, large valves are useful 
to allow by-passing the diffusion pumps. For example, in 
the vacuum system shown in Fig. 4, a large 4-inch valve 
makes it possible to open up the main vacuum chamber 



machine to 




*Y " bronze 

De fthotinsky 

and re-evacuate it without destroying the vacuum in the 
diffusion pumps. Valves between various parts of a large 
vacuum system facilitate narrowing the search for leaks, 
since one part after another can be isolated. 

The simplest valve for venting a vacuum system is a short 
length of rubber hose and a pinch clamp. Rubber vacuum 
hose is now available in sizes up to 1 inch in diameter. 39 
This large hose may be used in short lengths on the high- 
vacuum side of the diffusion 
pump when the pumps have a 
high capacity and when a vac- 
uum of only 10 -4 is desired. 
Usually, however, it is advis- 
able to confine the use of rub- 
ber hose to the low-vacuum 
side of the diffusion pumps. 
Ordinary plumbing valves 
can be modified for use in high- 
vacuum work. The glands are 
repacked with twine soaked 
in Apiezon compound a Q," 
beeswax, stopcock grease, or 
universal wax. Since the rub- 
ber gaskets supplied in these 
valves are often too hard for 
vacuum work, it is necessary 
to replace them with softer rubber. It is advisable to make a 
new end for the valve so that the new gasket rubber can be 
retained in a groove. The outside of the valve may be 
painted with shellac or glyptal lacquer as insurance against 
leaks, it may be coated with Apiezon wax "W," or it may be 
tinned. DuMond and Rose have described valves equipped 
with a sylphon bellows as a substitute for a packing gland. 40 
This is illustrated in Fig. 30. A packless valve of this type 

39 Small hose is obtainable from scientific supply houses. Large sizes of 
vacuum hose are sold by Central Scientific Company, Chicago, Illinois. 

40 DuMond, J. W. M., Rev. Sci. Instruments, 6, 285 (1935). 
Rose, John E., Rev. Sci. Instruments, 8, 130 (1937). 

valve 5 




Fig. 30. 



manufactured by the Hoffman Company can be readily 
adapted to vacuum work as shown in Fig. 31. 41 

Ordinary stopcocks can be sealed with stopcock grease 
for use in a high-vacuum system. Stopcock grease is made 
by digesting 1 part pale crepe 
rubber cut in small pieces with 
1 part Apiezon compound 
"M." This digestion is carried 
out in a balloon flask with pro- 
longed mechanical stirring at 
an elevated temperature ob- 
tained by means of a water or 
steam bath. 

When it is necessary to 
avoid grease on a stopcock, 
bankers' sealing wax, Apiezon 
wax "W," or Picein can be 
used. 42 Of these waxes, Picein 
exhibits the best body. With any one of them the valve 
is warmed until the wax becomes plastic each time that it 
is turned. (See Fig. 32.) Stopcocks may be lubricated 
with dry graphite and sealed with mercury. 

Mechanical motion. Mechan- 
ical motion can be introduced 
into a vacuum system through 
nonferrous vacuum walls with 
a magnet. An armature or bar 
magnet is fastened to the mov- 
ing part inside the system and 
actuated by an electromagnet 
outside. The armature can be 
hermetically sealed in a glass 
tube to avoid outgassing. 

Fig. 31. Hoffman packless valve. 


Fig. 32. Zaikowsky stopcock. 
U. S. Patent 2000552. 

41 Hoffman Specialty Company, Waterbury, Connecticut. Crane and Com- 
pany are local agents. 

42 For a description of a greaseless valve using a silver bellows acting against 
a silver chloride seat, see Ramsperger, Herman C, Rev. Set. Instruments, 2, 738 



A metal bellows can be used to introduce the reciprocating 
or oscillating motion of a lever. 43 When the end of the 
lever executes a circular motion, this motion can be trans- 
formed into rotation inside the vacuum. 

Van de Graaf has developed the high-speed sealed shaft 
shown in Fig. 33. The packing used is Apiezon grease "M" 
charged with graphite, and the pumping action of the right- 

and left-handed screws, cut 
on the shaft, prevents the 
extrusion of the packing com- 

Mechanical motion can be 
introduced through an ordi- 
nary packing gland packed 
with cotton twine <s&aked in 
Apiezon compound "Q" as 
shown in Fig. 4. 

Leaks. In planning a metal 
vacuum system, a part of the 
construction cost should be 
set aside to provide suitable 
fittings, plugs, plates, and tie bolts. The use of these makes 
it possible to pump air or hydrogen into separate compart- 
ments of the apparatus until the pressure is 50 or 100 
lbs. /square inch. For detecting leaks the pumped-up com- 
partment is submerged in water or painted with liquid- 
soap solution. Hydrogen, which may be used instead of 
air to pump up the apparatus, has the advantage over air 
that it diffuses through small holes approximately four 
times faster. When leaks are found, they may be repaired 
by welding or soldering or by merely peening the surface. 
After the whole apparatus is put together, the outside of the 
system is coated with several layers of glyptal varnish, al- 
ternating the color of the varnish coats, say blue and red, to 
facilitate complete coverage with each one of them. If possi- 
ble, the coating is baked at a temperature of about 120°C. 

43 Brose, H. L., and Keyston, J. E., Journ. Sci. Instruments, 7, 19 (1930). 


and colloidal graphite 

in packing gland 

Fig. 33. 



Leaks are usually found in a glass apparatus by passing 
the ungrounded high-potential electrode of a spark coil or 
high-frequency coil over the surface of the glass. When the 
electrode comes near the leaking channel, a spark jumps to 
it and causes residual gas inside the apparatus to become 
luminous. As a safety precaution, a spark gap of J to ^ inch 
should be connected in parallel with the electrode and the 
ground to prevent an excessive potential which might 
puncture the glass. 

Leaks in metal apparatus which are not detected by 
immersing the apparatus in water or painting it with soap 
solution are more difficult to locate. In general, the pro- 
cedure for finding them involves covering the walls of the 
apparatus with a liquid which solidifies, with water, or with 
a gas. In any case, while the search is in progress, the 
apparatus is maintained at the lowest pressure possible. 

If a liquid covering is used, it is, applied to local areas in 
progression until the offending region is located. As cover- 
ing one may use a molten mixture of beeswax and resin, or 
it may be a thick solution of either shellac in alcohol or 
glyptal lacquer brushed on the walls, or it may be cellulose 
acetate solution sprayed on the walls. When a solution of 
shellac (or lacquer) is applied to the outside of a leaking 
channel, the solution is drawn into the channel by the 
vacuum. As the solvent evaporates from this solution into 
the vacuum chamber, the liquid in the channel congeals. 
Thus, the leaking channel is, in effect, filled with a solid 
shellac core. The amount of solvent passing into the 
vacuum through this core is negligible in cases where the 
procedure is suitable. 

When the leak is covered with the solution, the vacuum 
usually improves at once. This improvement may be indi- 
cated by the disappearance of luminosity in a connected dis- 
charge tube and finally by sparking across an alternate gap. 
If an ionization or Pirani vacuum gauge is used, covering of 
the leak is indicated by motion of the spot of light on the 
scale of the instrument. 


The general region in which leaks are located may be 
determined by temporarily covering the region with water. 
As the vapor pressure of water is only about -^ of an atmos- 
phere, the leak may be expected to be attenuated 30-fold 
when it is covered. 

The third procedure for finding leaks involves covering 
general regions of the apparatus with gas, carbon dioxide 
for the top parts, since it is heavier than air, and illuminating 
gas for the bottom. Webster has described the use of a 
rubber "coffer dam" to facilitate the management of the 
gas. 44 Illuminating gas may be blown on various parts of 
the apparatus from a hose, or the surface may be gone over 
with a wad of cotton wet with ether. Evidence that the 
leak is admitting gas instead of air is a change in character 
of the luminescence in a discharge tube connected to the 
apparatus or a change in reading of a vacuum gauge sepa- 
rated from the apparatus by a liquid air trap. 

There are two procedures for using a discharge tube 
with illuminating gas, carbon dioxide, or ether. By the first, 
the obtainable vacuum is necessarily so poor, on account of 
the leak, that a distinct discharge is obtained. When the 
leak is covered, the luminosity in the positive column changes 
from the brownish-red color characteristic of air to the 
bluish-green of carbon dioxide or to the white of gas and 
ether. By the second procedure, used when the leak is 
small and a lower pressure is attainable in the system, the 
luminosity in the discharge is feeble. Webster suggests 
connecting the discharge tube behind one of the diffusion 
pumps as shown in Fig. 4. The backing pumps are then 
shut off, preferably just behind the discharge tube con- 
nection. The diffusion pump compresses the gas which the 
leak may be admitting, resulting in a more brilliant lumi- 
nescence in the discharge tube. 

A liquid air trap may be connected between the apparatus 
and a vacuum when carbon dioxide or other condensable 
gases are used. With this arrangement, when the leak is 
44 Webster, D. L., Rev. Sci. Instruments, 5, 42 (1934). 


admitting carbon dioxide, the trap condenses this gas, thus 
preventing it from entering the gauge. At the same time 
air and other gases which do not condense in the trap are 
removed by the pumps. As a result, even though the 
pressure in the system may have increased, an improvement 
of the vacuum is indicated. 

Obviously, a gauge which reads continuously (Knudsen, 
Pirani, or ionization gauge) is preferred to a McLeod 
gauge for hunting leaks. Relative rather than absolute 
readings of the pressure are sufficient for locating leaks. 
Thus, the Pirani and ionization gauges are satisfactory, 
although they do not give absolute pressure determinations. 

Vacuum gauges. A vacuum gauge determines the 
pressure in an evacuated apparatus by a measurement of 
some physical property of the residual gases, such as vis- 
cosity, heat conductivity, and so forth. The measurement 
of the response of a gauge to the residual gas naturally be- 
comes more delicate as the gas becomes more and more 
tenuous. Finally, below a certain pressure limit (which is 
characteristic of a given gauge) the gauge does not behave 
measurably different from what it would if the vacuum were 
perfect. For example, a discharge tube will give qualitative 
indications of pressure down to about 10~ 3 mm of mercury. 
Below this pressure the tube becomes nonluminous and non- 
conducting. The characteristic limits for some of the other 
gauges are as follows: 

Ionization gauge 10~ 9 mm of mercury 

Knudsen gauge 10~ 6 mm of mercury 

McLeod gauge 10~ 6 mm of mercury 

Pirani gauge 10 -5 mm of mercury 

Langmuir's viscosity gauge 10 -5 mm of mercury 

The operation of the McLeod gauge depends on a definite 
volume of residual gases being compressed, so that as the 
volume decreases, the pressure is increased to a value at 
which the hydrostatic head of mercury can be measured 
with an ordinary scale. 

The ionization gauge measures with a galvanometer the 
positive ions that are formed in an electric field when the 


residual gas is bombarded with electrons. The Langmuir 
gauge depends on the measurement of viscosity, and the 
Pirani gauge on the measurement of heat conduction of the 
residual gas. The Knudsen absolute manometer measures 
the momentum transferred from a hot to a cold surface by 
the gas molecules. 

Of the above gauges, only the McLeod and Knudsen are 
absolute manometers in the sense that their geometry and 
other measurable characteristics of construction and opera- 
tion determine their response at a given pressure. The 
McLeod gauge is the simplest and most reliable for perma- 
nent gases, but it has the disadvantage of giving erratic 
response or no response at all to water vapor, carbon dioxide, 
ammonia, and pump oil vapors which adsorb on the walls of 
the gauge or condense to a liquid. This disadvantage is 
serious, inasmuch as water vapor, carbon dioxide, and 
so forth are often of importance in the last stages of ob- 
taining a high vacuum. The Knudsen gauge responds to 
gases and vapors alike. 

The response of an ionization gauge is difficult to predict 
from its construction details, and it must be calibrated with a 
McLeod gauge using permanent gases. Furthermore, be- 
fore the pressure can be inferred, it is necessary to make 
corrections for the molecular weight of the gas and also for 
the possibility that the gas may be dissociated by the elec- 
tron bombardment. Quantitative application of the gauge 
is unreliable to the degree to which these corrections are 
uncertain. Likewise, the response of the Pirani gauge de- 
pends on the molecular weight of the residual gas, and it 
must be calibrated with a McLeod gauge that uses perma- 
nent gases. The same is true for the viscosity gauge. 

The McLeod gauge. 45 Although many improvements 
have been made in the McLeod gauge, they have seldom been 
applied. The gauge as ordinarily used today is essentially 

^Gaede, W\, Ann. d. Physik, 41, 289 (1913). 
Hickman, K. C. D., J.O.S.A., 18, 305 (1929). 
Pfund, A. H., Phys. Rev., 18, 78 (1921). 



the same as it was originally. We will discuss here the 
simple form of the gauge illustrated in Fig. 34. It is made 
of glass as shown and is mounted on a vertical board. The 
difference in the heights of the mercury levels in the gauge 
and in the reservoir is approximately equal to the barometric 
pressure B. As the reservoir is raised, the mercury level in 
the gauge comes above the Y-branch, thus isolating a definite 
volume Vi of the residual gas. This is isolated at the un- 
known pressure P h the pressure of the residual gas in the 
apparatus to which the gauge is connected. As the mercury 
reservoir is further raised, the isolated residual gas is com- 
pressed, and when its volume has been reduced to a volume 
V 2j the pressure is great enough to produce a sensible 
difference in the height of the mercury meniscus in the two 
capillaries, A and B. At the left, in Fig. 34, the mercury 
levels are shown at the beginning of a measurement, and at 
the right they are shown in two different positions corre- 
sponding to two methods of making readings. In one, if 
the meniscus in B is adjusted to the same height as the top 
of capillary A, the final volume, V 2 , is equal to Ah -a, when 
<t is the cross-section area of the capillary. The decrease 
in volume from Vi to V 2 is ordinarily of the order of one- 
hundred-thousandfold, with a corresponding increase of 
pressure in the capillary over that which obtained originally. 
The construction of the gauge with the comparison capillary 
B of identical bore with A eliminates the necessity of mak- 
ing corrections for surface tension. Referring to Eq. 1, we 
see that the product PiVi is, in this case, a constant. The 
original product, PiKi, is equal to the final product, P 2 F 2 . 
From this we get the expression connecting the unknown 
pressure with the observed manometer difference, Ah: 

ft = ^- d4) 

Vi and a are constants of the gauge determined when it is 
constructed. <j is obtained by measuring the length of a 
known volume or weight of mercury in the capillary. Y\ is 


determined by filling the gauge with mercury. These 
original data may be recorded on the board to which the 
gauge is attached. Here they will not be lost. Values of 
Pi determined by Eq. 14 are usually laid off on a nonlinear 
scale, which is mounted behind capillary A in order that 
pressures may be read directly. 

The second procedure of making the observations on V2 
and P 2 is illustrated at the right in Fig. 34. The gas is com- 
pressed to a definite mark on capillary A at a distance Ah 
from the top, so that the final volume, V2, is the same for 
every measurement. The final pressure necessary to com- 
press volume Vi to V2 is Ah, and the pressure Pi in the sys- 
tem is determined by these quantities, according to the 
following equation: 

P. = ?pAh. (15) 

A linear pressure scale computed from this formula is 
ordinarily mounted behind capillary B. 

The McLeod gauge is thoroughly reliable for the perma- 
nent gases from 10 -1 mm to 10~ 4 mm of mercury. It is 
less reliable to 10 -5 mm. Below this the indications are 
only qualitative, and at 1CT 6 the mercury often sticks in the 
top of capillary A. 

The gauge is most reliable after it has been outgassed by 
gently warming it with a soft flame. Three gauges with 
different values of Vi are necessary to cover adequately the 
complete pressure range from 10 -1 to 10~ 6 mm. Many of 
the designs of McLeod gauges are more elaborate than the 
one shown in Fig. 34. For example, three bulbs may be 
mounted together with one reservoir, one for low pres- 
sures, one for intermediate pressures, and one for high 

The McLeod gauge is fragile. If it breaks, not only is 
the gauge lost but what is often more serious, mercury may 
get into the vacuum system. In glass vacuum systems using 
mercury pumps this is not as serious as it may be in kinetic 
vacuum systems. These systems, fabricated of brass with 


soft-soldered joints, are attacked by mercury and the joints 
are destroyed. 

Accidents with this gauge are usually caused by bringing 
the reservoir up too quickly. Then mercury in Vi acquires 
enough momentum to shatter the bulb when the metal 
surface arrives at the opening of the capillary tube with no 
cushion of air to soften the shock. 

Admitting air into the vacuum system is to be avoided 
when the mercury is not completely out of Vi. The ad- 
mission of air will have the same result as carelessness in 
raising the reservoir. 

Sometimes a mercury pellet will remain in capillary A 
when the reservoir is lowered. It can usually be brought 
down by tapping the capillary (after the mercury is all 
out of Vi). If this treatment fails, the capillary should be 
heated with a soft gas flame. In the latter case, a sheet of 
asbestos is placed behind the capillary to protect the cali- 
bration scale from the flame. 

The capillary tubes used for the construction of McLeod 
gauges are seldom larger than 2 or 3 mm or smaller than 
\ mm bore. The volume of the bulb, Vi, ordinarily varies 
from 50 to 500 cc. Only pure distilled mercury should be 
used. Mercury is attacked by the sulphur present in rubber 
hose, so that dross is produced which adheres to the inside 
of the gauge and may become very annoying. A gauge 
contaminated with this sulphide may be cleaned out by 
the combined action of zinc dust and nitric acid. Rubber 
hose for use on a gauge should be cleaned before it is used 
by passing hot caustic potash solution back and forth 
through it for a quarter of an hour or so. The tubing 
should be thoroughly washed free of caustic and dried 
before use. 

In cases where it is necessary to avoid contamination of 
the vacuum system with mercury vapor, a liquid air trap 
should be connected between the vacuum system and the 
gauge. For kinetic vacuum systems this precaution is 
often omitted. A stopcock between the gauge and the 




3 hon t rt^r^,/ 

system which is kept closed when the gauge is not in use 
minimizes contamination. 

The ionization gauge. 46 Ionization gauges are triodes 
mounted in a glass bulb connected to the apparatus in which 
the pressure is to be measured. They are electrically con- 
nected as shown in Fig. 35. 

Electrons emitted from the filament are accelerated to 
the grid, and their momentum would carry them to the 
plate if an inverse field more than sufficient to prevent this 
were not impressed between the 
grid and the plate. They there- 
fore return to the grid and are 
finally collected on it. However, 
while they are between the grid 
and the plate, they bombard and 
ionize some of the molecules of 
the residual gas present there. 
These ions are collected on the 
plate and measured with a sensi- 
tive galvanometer. The ratio 

of this ion current to the current of bombarding electrons or 
grid current is proportional to the pressure at pressures below 
about 10 -4 mm. 

An ionization gauge may be made from an ordinary 
three-element radio tube equipped with a glass connection 
to the vacuum system. Such gauges are useful for the 
pressure range from 10~ 3 to 10 -6 mm of mercury. 

Fig. 36 shows the construction details of a gauge designed 
to have higher insulation of the plate than an ordinary radio 
tube. Measurements with it are possible to a pressure 
of 10 -9 mm of mercury. The upper end of a glass bulb 
supports the plate assembly, while the lower end supports 
the combined grid and filament assembly. The grid is 

46 Buckley, O. E., Nat. Acad. Sci., Proc, 2, 683 (1916). 
Dushman, S., and Found, C. G., Phys. Rev., 17, 7 (1921). 
Jaycox, E. K., and Weinhart, H. W., Rev. Sci. Instruments, 2, 401 
Simon, H., Zeits. f. techn. Physik, 5, 221 (1924). 

Fig. 35. 










made from a piece of nickel screen rolled to form a cylinder. 
This is bound mechanically to the central glass tube through 
the bottom by wrapping it with wire, and it is connected 
electrically to the grid electrode with one loose end of the 
wrapping wire. There are two filaments, but only one is 
used. The other is held in reserve to be used if the first is 

accidentally burned out. The 
filaments may be replaced by 
cutting the central tube at S. 
Expensive auxiliary electri- 
cal instruments are required 
for this gauge. They should 
be protected with Littelfuses 
as shown in the wiring diagram 
(Fig. 35). 

The plate may be outgassed 
with high-frequency currents 
or by electron bombardment. 
In the latter case, an alter- 
nating potential of 500 volts is 
applied between the filaments 
and the plate. The amount of 
heat developed depends on the 
emission from the filament, 
and this is controlled by the 
filament current. Outgassing 
of the plate and glass 
walls of the gauge is necessary if quantitative measurements 
are to be made. However, for hunting leaks it is necessary 
only to out gas the plate once. 

Dunnington has made a gauge using 30-mil helices of 
tungsten wire for both plate and grid. These helices are 
outgassed simply by passing a current through them for a 
few seconds. He found that such a gauge did not have a 
linear relationship between pressure and ratio of plate to 
grid currents. Once calibrated, however, it was found to be 
very reliable. 


to vacuum 

Fig. 36. 



At a given pressure, the ratio of plate to grid current is 
different for different values of the grid current. For this 
reason, it is necessary to adjust the grid current to some 
definite value, usually in the range of 10 to 50 milliamperes. 

The Pirani gauge. 47 The Pirani gauge consists of a heated 
filament of platinum, tungsten, or some other metal with a 
high temperature coefficient of electrical resistance. The 
filament is exposed to the residual gases and is cooled by 
them. The temperature of the filament is determined by 
the heat conductivity of the residual gas, which, in turn, 
depends on the pressure. The filament may be operated 
in several ways. The most 

cter *en<si 

Fig. 37. 

satisfactory method is to con- 
nect the filament to one arm 
of a Wheatstone bridge and 
heat it by a constant current 
as shown in Fig. 37. If the 
bridge is balanced at one tem- 
perature of the filament, a 
change of its temperature 
caused by a change in the 

heat conductivity of the residual gases will unbalance it. 
Thus, the deflection of the bridge galvanometer indicates the 
pressure of the residual gases. 

Ordinarily, the filament is mounted in a bulb fitted with a 
connecting tube and is balanced with an identical compen- 
sating filament mounted in an adjacent arm of the bridge. 
This auxiliary bulb is evacuated and sealed off at a very low 
pressure. The use of an auxiliary bulb serves to make the 
gauge insensitive to variations in room temperature. 
Changes in the over-all temperature of one bulb are the same 
as changes in the other, so that the galvanometer does not 

47 DuMond, J. W. M., and Pickels, W. M., Jr., Rev. Sci. Instruments, 6, 
362 (1936). 

Hale, C. F., Am. Electrochem. Soc, Trans., 20, 243 (1911). 
von Pirani, M., Deutsch. Phys. Gesell, Verh., 8", 24 (1906). 
Skellett, A. M., J.O.S.A., 15, 56 (1927). 
Stanley, L. F., Phys. Soc, Proc, 33, 287 (1921). 






to 3 





deflection of galvanometer* 
Fig. 38. 

respond to these changes but only to the changes produced 
by the residual gas in the one bulb. 

Fig. 38 shows a calibration 
curve of a Pirani gauge manu- 
factured by E. Leybold Nach- 
folger. The pressure range 
over which it is useful extends 
from xV mm to 10~ 4 mm. 

The construction of the Pi- 
rani gauge, together with the 
theory of its use, is treated in 
detail by several authors, who 
should be consulted by anyone 
planning to use the gauge for quantitative measurement. A 
gauge useful for qualitative work, as for hunting leaks, can 
be improvised from two 
ordinary 20- to 40-watt vac- 
uum tungsten lamps, one of 
which is fitted with a con- 
necting tube. Fig. 39 shows 
the construction details for 
this gauge. The bridge gal- 
vanometer should have a 
sensitivity of about 10 -8 
ampere division. Sometimes 
uncertain contact to the sup- 
porting wires may cause 
variable heat loss from the 
filament, and this should be 
suspected if the gauge is 
erratic. Tapping will often 
define the contact. 

The Langmuir gauge. 48 
Langmuir 's viscosity gauge is Fig. 39. 


P 6 



. cable to 

© auxil iary 













8 •St' - 







Pirani lamp K. 

permanently evacuated at 10" 6 mm 

48 Beckman, Arnold O., J.O.S.A., 16, 276 (1928). 
Haber, F., and Kerschbaum, F., Zeits. f. Elektrochem. 
Langmuir, I., Am. Chem. Soc, J., 35, 107 (1913). 

296 (1914). 




made with a flattened quartz fiber about 50/z thick and from 
five to ten times as wide. This quartz ribbon is about 
5 cm long and is mounted in one end of a glass tube about 
25 mm in diameter, as shown in Fig. 40. When this ribbon 
is set vibrating in a high vac- 
uum, the amplitude changes 
very slowly because the damp- 
ing by the residual gas is almost 
negligible, and, owing to the 
low internal viscosity of fused 
quartz, the loss of vibrational 
energy from this source is 
also low. From atmospheric 
pressure down to a few milli- 
meters of mercury, the damp- 
ing produced by the molecules 
of the residual gas is nearly in- 
dependent of pressure. Over 
the transition range of pres- 
sure, where the damping varies 
from this constant value to 
zero, the time required for the 
amplitude of vibration to de- 
crease to half value is an index 
of the pressure. Within this range the relation between the 
time, t, the pressure, P, and the molecular weight of the 
residual gas is given by the following formula: 

iron arma- 
ture to 
P activate 
- — in glass 

to- vacuum 

Fig. 40. 

PVM = - t -b. 


Here a and b are constants of the gauge. The value of the 
ratio b/a may be obtained by observing the damping time, 
t , for an essentially perfect vacuum, that is, a pressure of 
10~ 6 mm or less. For this pressure the left side of Eq. 16 
can be set equal to zero. The values of a and b are deter- 
mined from a second measurement of the time U at a definite 
pressure Pi. This pressure is determined with a McLeod 



gauge. M is approximately 29 for air. The gauge may also 
be calibrated by subjecting it to saturated mercury vapor 
at a definite temperature at which the vapor pressure 
of mercury is known. The range over which the gauge is 
most useful lies between the pressures 2 X 10 _2 and5 X 10~ 5 . 
A feature of this gauge is its small volume. Because there 
are no metal parts exposed, the gauge is suitable for measur- 
ing the pressure of corrosive gases like the halogens. This 
gauge, in conjunction with a McLeod gauge, may be used 

for measuring the molecular 

weight of an unknown gas at 
low pressures. 

The flat quartz fibers may 
be obtained by drawing them 
out of the side rather than 
the end of a quartz tube or by 
following the technique given 
in Chapter V. 

Figs. 40 and 41 show con- 
struction details and the 
method of mounting the fiber 
together with a pivoted glass 
tube, which contains an iron 
armature operated by an external electromagnet, to start the 
fiber vibrating. An optical arrangement for observing the 
amplitude of vibration is also shown. An image of the 
quartz fiber is projected on a scale with a simple lens. 

The Knudsen gauge. 49 Fig. 42 shows the Knudsen 
gauge as designed by DuMond. When this gauge is con- 
structed according to the specifications outlined by him, it 
is claimed to have a definite sensitivity, so that no prelimi- 
nary McLeod calibration for it is needed. The gauge shown 
here differs slightly from DuMond' s design in that it is 
equipped with a permanent (Alnico) magnet for damping. 

Fig. 41. 

49 DuMond, J. W. M., and Pickels, W. M., Jr., Rev. Set. Instruments, 6, 362 
(1936). ' 
Knudsen, M., Ann. d. Physik, 28, 75 (1909). 

taper pin foip^ 
of vane 

^ liquid ait* trap 



lens cemented on window; 

2.5- mil aluminum vane 
32-mil aluminum-wireaxle 
threaded through vane and 
spot welded 
galvanometer mirror 
Vz mil tungsten-wire suspen- 
sion, free length- 5 cm 
wire eyes for safety 
X M" copper rods 
heaters - 5 feet of No.£8 
Chromel wire, £5 ft each 
insulated leads to heaters 
permanent magnet for 
pole-pieces - iron 
water jacket 
13 straight filament lamp 
scale of inches for plan and section 

Fig. 42. 


Also, it has a special liquid air trap for determining what 
fraction of the pressure indication is produced by con- 
densable vapors. 

The Knudsen gauge is to be preferred to the McLeod 
gauge where it is important to avoid contaminating a vacuum 
system with mercury. No expensive auxiliary instruments 
are required with the Knudsen gauge, as with the ionization 
gauge. Furthermore, the filaments will not burn out and 
the suspension is not delicate. 

It is advisable to modify DuMond's design so that all 
connections and supports fasten to one end plate. This 
facilitates making repairs. The metal case thus becomes, 
in effect, a water-cooled covering "bell jar" fitted with a 


Coating of Surfaces: Evaporation 
and Sputtering 

GLASS, quartz, and other nonmetallic substances may- 
be coated in the laboratory with thin films of metal by 
the following processes : 

1. Burning on 

2. Chemical deposition 

3. Cathode sputtering 

4. Evaporation 

Each of these is characterized by certain restrictions and 
advantages. For example, the "burning-on" method is 
applicable only in cases where the glass can be heated; 
chemical silvering (and also coating with gold and copper 
from aqueous solution) cannot be applied to surfaces like 
rock salt which are attacked by water; sputtering is par- 
ticularly suitable for preparing films of the platinum metals; 
and the evaporation process is suited to the application of 
aluminum films. 

Although deposits can be produced on metals as well as 
nonmetals by these processes, electroplating (not treated 
here) is usually the most practical for coating metals. 

Burning-on method. Glass may be coated with a thin 
film of metal by the burning-on process. The process is 
applicable for the noble metals, which are reduced by 
heating. The glass to be coated is covered with a layer of 
an oily solution of one of the metallic salts. When heat is 
applied, the oil burns away, and the salt is reduced, leaving a 
deposit of the metal. This deposit is formed in an adherent 



compact film by a final heating to the softening point of the 

A solution for depositing platinum 1 is made as follows: 
Evaporate 100 cc of a 10 per cent H 2 PtCl 6 solution to 
dryness and dissolve it in a minimum quantity of absolute 
alcohol. Add this alcohol solution slowly to 6 cc of oil of 
lavender kept ice-cold. Finally, add some Burgundy pitch 
to give the mixture consistency, so that it will remain uniform 
when it is applied and the glass is slowly heated. 

Solutions for gold, silver, and iridium are available 

A platinum film burned onto porcelain may be electro- 
plated with copper and soldered, thus affording a method of 
making a vacuum-tight seal between metal and porcelain. 

Chemical silvering processes. 2 There are two widely 
used methods for chemical silvering. These are the Brashear 
method and the Rochelle salt method. The first is used to 
obtain thick coats on front-silvered mirrors which are to be 
frequently burnished, such as telescope mirrors. The Ro- 
chelle salt method, because its action is slower, is recom- 
mended for making partially silvered mirrors, such as inter- 
ferometer plates, which require a uniform thin film with a 
specified ratio of reflection and transmission. 

Cleaning. The silver film does not deposit well on con- 
taminated surfaces. Therefore, fats and other surface 
contaminations must be cleaned off the glass, so that the 
colloidal particles of silver suspended in the silvering solu- 
tion will adhere strongly to the glass to form a tenacious 
compact metallic film. Just as a greasy glass surface is 
difficult to wet with water, so a clean wet surface does not 

1 McKelvy, E. C, and Taylor, C. S., "Glass to Metal Joints," Amer. Chem. 
Soc, J., 42, 1364 (1920). 

2 Gardner, I. C., and Case, F. A., "The Making of Mirrors by the Deposition 
of Metal on Glass," Bureau of Standards Circular No. 389. 

Ingalls, Albert G., editor, Amateur Telescope Making. New York: Scientific 
American Publishing Company, 1935. 

"The Making of Reflecting Surfaces," a discussion held by the Physical 
Society of London and the Optical Society, November 26, 1920. London: 
The Fleetway Press, Ltd. 


readily take up greases, fats, and other contamination. 
Accordingly, once a surface is clean, it will stay clean, if it 
is kept under distilled water until it is immersed in the 
silvering solution. 

The first step in cleaning a mirror is to free the sides and 
back of it from rouge and all other contaminations. An ink 
eraser is ideal for the removal of such contaminations. The 
pumice or ground glass in the eraser has an abrasive action 
particularly suitable for this preliminary cleaning of non- 
optical surfaces. The polished face cannot be cleaned in 
this manner, but it is well to work the eraser well over the 

The mirror is next washed all over with soap and water, 
or Aerosol 3 and water. Aerosol is preferred to soap, since 
it may be washed off the face of the mirror without leaving 
any residue. If soap is used, it should be rinsed off with 
rain water or, better yet, with distilled water. 

A mild and harmless abrasive action on the face of a mirror 
is sometimes necessary. This is obtained by rubbing it 
with a pad of wet cotton, to which some precipitated chalk is 
added. After a polished glass surface has been treated with 
chalk, the cleaning water should wet the whole mirror face 
and not draw back anywhere to leave dry areas. It may be 
necessary to repeat the chalk treatment several times. 

The mirror is next rinsed with water and swabbed with 
concentrated nitric acid, a powerful oxidizing agent which 
removes organic matter adsorbed on the glass. The swab 
for applying the acid is made by wrapping absorbent cotton 

8 The compound Aerosol OT is manufactured by the Selden Division of the 
American Cyanamid and Chemical Corporation, Bridgeville (Pittsburgh), 

Duncan, R. A., Indust. and Engin. Chem., 26, 24 (1934). This article gives 
a description of new detergents of which Aerosol is an example. These deter- 
gents have in common the constitution of sulphonated organic compounds of 
high molecular weight. They have a neutral reaction, and their advantage 
iver soap for washing mirrors lies in the fact that they may be used in neutral, 
caustic, or even acid solutions. Unlike soap, they form soluble compounds 
with magnesium and calcium ions, which are common in tap water. The 
detergent Dreft, obtainable at grocery stores, is also suitable for washing 


on a glass rod and fastening it with cotton twine as shown in 
Fig. 1. Care is exercised in using the swab to prevent the 
end of the rod from coming in contact with the mirror face. 
This nitric acid treatment should be carried out in the con- 
tainer in which the mirror is to be silvered to avoid possible 
contamination later with oil from the hands when the mirror 
is handled. If it is necessary to handle the mirror, it is 
advisable to use rubber gloves. 

Cleaning solution (chromic and sulphuric acid mixture) 

may be used for cleaning glass, but it is not ordinarily 

string >-■ ^ r " = T"~^\ necessary. This solvent is very 

effective. Even paraffin and 
carbonized organic material 
may be removed from glass in 

ftiS!ng •'od/J^^ 8 ^^ cases where the glass and the 
_. , chromic acid solution can be 

Fig. 1. 

heated together. 

After being rinsed with tap water, the mirror is treated 
with a concentrated solution of stannous chloride. This is 
removed after a few minutes by a very thorough rinsing. 
All chloride ions must be washed away, first with tap water 
and finally with distilled water. The mirror can stand in 
the distilled water until the silvering begins. 

It is important in silvering to clean carefully all the recep- 
tacles and graduates used. A long stick with an ink eraser 
fastened to the end will be found helpful to remove water 
stains and other contaminations. 

Brashear's process. 4 The Brashear process is described 
graphically in Fig. 2. The three formulas for the reducing 
solution given there afford different ways to effect the same 
end. In the first formula the nitric acid slowly digests the 
table sugar, to yield the sugars dextrose and levulose. 
This requires time and so the solution must be aged before 
use. In the second formula this aging is accelerated by 
boiling, and the solution can be used as soon as it is cool. 

4 Brashear, John A,, English Mechanic, 31, 237 (1880). 
Wadsworth, F. L. 0., Astrophys. J., 1, 352 (1895). 

.Stock Solutions 



300 cc water Ammonia 100 cc water 30 cc water L- — -J 
20g AgN0 3 (concentrated) I4g ROH 2g AgN0 3 J""^ 

1 liter water 

All water used must be distilled 
1 st step 
n Pour ammonia into solu- 
/ tion A until a darK brown 
' precipitate of silver oxide 
forms and begins to clear. 

2nd step 
Add ammonia drop, by 
drop until solution just 

| J clears. Disregard small 

F-— A specks. Stir well between 
NH+OH drops near the end 

90 g table sugar 
4cc HN0 3 (cone) 
175 cc Alcohol 
Age for 2 weeKs 
or a month. 

120 cc water 
11 g table sugar 
4.SccHN0 3 (conc) 
Boil and cool. 
Ready for use 

120 cc water 
7.8 g dextrose 
Ready for use 

3rd step 
Add solution C (AgN0 3 ) 
drop by drop until the 
solution is a distinct 
straw color. This is to 
avoid excess ammonia. 

4th step 

CAUTION From here on there is danger of 
an explosion Use goggles. 

Add all of solutiorfB slowl y and stir 
constantl y. 

5th step 
Add ammonia, a dropper full at a time 
and finally a drop atatime, until the 
solution just clears. 

6th step 

Add solution C drop by drop until there 
is a thin straw-colored or brownish pre- 
cipitate. (Disregard small specKs.) 

7 th step 
Filter through cotton. 

Reducer tast * t6 P 
1 I.ILorlD Add 120 cc of reducing 
-* — £ solution and Dour immedra 

...... pour immediately 

over the mirror to be silvered. 



Fig. 2. 


In the third formula dextrose is used directly. The alcohol 
is a preservative, and it is not required for the second and 
third solutions unless they are to be stored, in which case 
the same proportion of alcohol is used as called for in the 
first formula. 

There is danger of an explosion after the fourth stage, in- 
dicated in Fig. 2. The formation of the explosive, fulminat- 
ing silver, is not particularly favored by the low concentration 
of solutions and moderate temperatures that obtain here, 
but these relatively weak solutions will give fulminate on 
warm days if they are allowed to stand. This compound ex- 
plodes on the slightest provocation when dry and sometimes 
when wet. Accordingly, all spent silver solutions should be 
rinsed down the sink at once. Goggles are recommended 
for safety. 

As soon as the reducing reagent is added, the silvering 
solution is poured over the mirror. Filtering is optional. 
The distilled water in which the mirror has been standing 
may or may not be poured off first. Soon after the reducer 
is added, the solution becomes dark brown and then black. 
After this, it gradually develops a muddy brown appearance. 
At this stage the deposit of silver on the mirror is already 
continuous or should soon become so. The container for 
the mirror and solution may be tipped from time to time to 
stir the solution and allow inspection of the surface. When 
the silver film covers the whole surface and as soon as black 
specks begin to settle on it, a light swabbing with a cotton 
pad is recommended. This rubbing must be delicate at first, 
but it may be more vigorous as the silver becomes thicker, 
the surface being inspected from time to time for bloom. 
Usually when the solution begins to clear, it is nearly spent, 
and since the possibility of bloom becomes greater at this 
stage, it is best to pour off the solution and rinse the mirror 
with distilled water. For a full silver coat, Brashear's pro- 
cess requires, on the average, from 6 to 10 minutes. 

If a bright light, such as the sun, is visible through the 
coat, it is too thin. In this case the mirror should be covered 


with distilled water, and the chemical solution for a second 
coat prepared. Do not let the mirror dry between coats. 

After a satisfactory coat is obtained, the rinsed mirror is 
rubbed with a pad of cotton until it is dry. The silver 
is burnished with a burnishing pad (chamois skin tacked on 
a Shinola shoe-polishing pad) to "compact" the coat. It is 
then polished with a similar chamois pad charged with 
optical rouge. The rouge pad may also be used from time 
to time to burnish away tarnish which forms on the silver 

Rochelle salt process. 5 Two solutions are required for 
the Rochelle salt process. Solution A is made as follows: 
5 g of silver nitrate are dissolved in 300 cc of water and 
ammoniated, as in the Brashear process, so that the silver 
oxide precipitate formed at first is almost but not completely 
clear. In case it inadvertently becomes clear, it must be 
back-titrated with a dilute solution of silver nitrate, so that 
the liquid finally presents a distinct straw color. This is 
filtered and diluted with water to 500 cc. Solution B is 
made as follows : 1 g of silver nitrate is dissolved in 500 cc 
of water. It is then brought to a boil, and 0.83 g of Rochelle 
salt, dissolved in a little water, is added. The boiling is con- 
tinued until a gray precipitate is deposited. The solution 
is filtered hot and diluted to 500 cc. These solutions may 
be stored for a month or so if they are protected from light. 

To silver a mirror, solutions A and B are mixed, volume 
for volume, and poured at once into the silvering vessel. 
The quantity of solutions given above is sufficient for a 
thick film on a glass surface of 200 cm 2 area. The tempera- 
ture recommended for silvering is 20°C. (68°F.). 

Silver is deposited slowly by the Rochelle salt process; an 
hour may be required for a thick deposit to form. Partial 
reflecting films are obtained as desired by withdrawing the 
glass from the solution at the appropriate time. The 
progress of the deposition may be judged from auxiliary 

5 This treatment follows that given in Miller, Dayton Clarence, Laboratory 
Physics, page 269. Boston: Ginn and Company, 1903. 


glass plates, which are removed from time to time to deter- 
mine the progress of the coating on the main plates. Fig. 3 
illustrates a simple test for determining when the silver film 
is half -reflecting (for 45° incidence). 

Partial reflecting plates are washed with distilled water 
and dried. Afterward they are polished by a light brushing 
with an eiderdown powder puff charged with optical rouge, 
as recommended by Pfund. 

Silver films are protected from tarnishing by covers of 
filter paper that have been soaked with lead acetate solution 

Light source should be directly over edge of mirror 

• black velvet V white paper 

less than half -silvered half-silvered rnorethan half-silvered 

Fig. 3. 

and dried. These covers are applied whenever the films are 
not actually in use. 

Lacquering. Another procedure for protecting the silver 
from tarnishing involves coating the film with a thin layer 
of colorless lacquer. The layer of lacquer destroys some of 
the reflectivity of the mirror, and in addition it exhibits 
interference colors. It. W. Wood has pointed out that a 
thin transparent film of lacquer on a good reflector should 
not show interference colors. 6 The colors usually exhibited 
by a lacquer film are due to frilling. This frilling can be 
observed directly only with the highest-power microscopes. 
Wood states that no frilling occurs and that there are, 
accordingly, no interference colors if collodion dissolved in 
chemically pure redistilled ether is used to lacquer the mirror. 

6 Wood, Robert W., Physical Optics, Third Edition. New York: The 
Macmillan Company, 1934. 


In order to obtain uniform lacquer films with the ether 
solution of collodion, it is necessary that the ether evaporate 
slowly. The can illustrated in Fig. 4 is suggested for use in 
lacquering with an ether solution. 

metal film can 
or candy box 

edges of mirror 
may be pro- ' 
tected with 
a rubber 

holes just below 
edge of cover 

15* Cover mirror 
with ether- collo- 
dion solution and 
close cover tightly. 

2n£ p our ou t 
solution without 
opening container. 

Fig. 4. 

3*^ Rotate con- 
tainer slowly while 
in a diagonal 
position until 
mirror is dry. 

Gold and copper. A chemical process for depositing 
gold from solution is described by von Angerer. 7 A process 
for copper is described by French. 8 

Sputtering. Although the sputtering phenomenon at 
the cathode of a glow discharge has been known for a long 
time, 9 the mechanism of the process is not fully understood 
even now. 10 There are two current theories of sputtering. 
One of these holds that the emission of metal by the cathode 
is pure thermal evaporation due to high temperatures at- 
tained in areas of molecular dimensions. These tempera- 

7 von Angerer, Ernst, Wien-Harms, Handb. der Exp. Physik, 1, 375 (1926). 

8 French, E. A. H., Optical Soc., Trans., 25, 229 (1924). 

9 Grove discovered the sputtering phenomenon in 1852. Grove, W. R., 
Phil. Trans., 1 (1852). 

10 Compton, Karl T., and Langmuir, Irving, Rev. Modern Physics, 2, 186 

Fruth, H. F., Physics, 2, 286 (1932), gives a comprehensive bibliography of 
cathode sputtering. 

Mierdel, G., Wien-Harms, Handb. der Exp. Physik, 13, Part 3, page 400 
et seq. (1929). 


tures are produced by the energy of impinging ions. The 
other theory invokes a mechanism for transferring the energy 
of the gas ion into energy of a metal molecule which is 
similar to the mechanism by which the energy of a light 
quantum is transformed to energy of an emitted electron. 
However, in spite of its being incompletely explained, 
sputtering is understood empirically, and its practical appli- 
cation for obtaining metal films on glass is simple. 

Sputtering can be carried out successfully under a wide 
variety of conditions. For example, the pressure of the 
glow discharge may range from 1 down to 10 -2 mm. The 
cathode is naturally made of the metal to be sputtered, al- 
though its shape may vary considerably. The anode is 
usually aluminum or iron. The glow discharge is prefer- 
ably produced by a direct potential, although an alternating 
potential can be used. The potential usually ranges above 
1000 volts and frequently is as high as 20,000 volts. The 
residual gas in the sputtering chamber may be air, hydrogen, 
argon, or other gases. (The sputtering rate with helium is 
extremely low, and this gas is used for glow discharges where 
sputtering is to be avoided.) The surface to be sputtered 
is usually placed tangent to the boundary of the cathode 
dark space, although it may lie within or beyond it. The 
low pressure required can be obtained with a mechanical 
pump of small capacity on a tight system or with a faster 
mechanical or diffusion pump on a system equipped with a 
regulating leak. 

A typical setup for the sputtering process is shown in 
Fig. 5. The sputtering chamber is usually a glass bell jar 
with a hole in the top for the cathode connection. It may 
be made from an old bottle with the bottom cut out and the 
base ground flat. It is best to have a glass plate for the 
base, although a metal one (preferably iron) will suffice. An 
aluminum plate can be used to cover any exposed metal 
parts which may give trouble by sputtering. It is advisable 
to heat all aluminum before it is used in order to drive off the 
machine oils which may be contained in it. Glass cylinders 



hard wax seal 

beeswax and ^> 
resin seal 


glass tube 

bell jar or 
bottle with 

and resin 

metal tube 

gas inlet 

to power 
looo to 2o,ooo volts 

cathode of 
metal to be 



metal or 
base plate 

to power supply 
1000 to 3o,ooo volts 

to vacuum pumps 

Fig. 5. 


and plates, as shown in Fig. 5, are useful for confining the 
discharge. If these plates and cylinders are not used, the 
outgassing induced by the discharge may give rise to foreign 
substances deleterious to the film produced. 

The cathode is fitted in the top of the bell jar as shown. 
It is pulled up against the square end of the depending glass 
tube by the connector wire. This wire is secured by wrap- 
ping it around the top end of the tube, where it is sealed with 
wax (Apiezon "W," shellac, or DeKhotinsky wax). 

Batteries or motor-generator sets are ideal sources for 
the sputtering potential, but other sources of potential are 
often employed. An induction coil makes a convenient 
source of potential, giving partially rectified current. How 
ever, alternating current from a 10,000-volt neon-sign 
transformer can be used. It is advisable but not necessary 
to rectify the current from this transformer with a Kenetron 

The use of a milliammeter to measure the discharge 
current is advisable when making partially transmitting 
coats. When the sputtering equipment has been calibrated, 
this current serves as an index to determine proper exposure 
for obtaining a desired ratio of transmission and reflection. 
The sputtering rate can be controlled, for example, by ad- 
justing the filament current of the Kenetron. The rate of 
sputtering increases a little more than linearly with the 
sputtering current, depending somewhat upon the condi- 
tions of temperature, pressure, and geometry which obtain. 
For work in which high reproducibility in the film thickness 
is required, it is advisable to use a fast pump and to wash the 
bell jar continuously with air or hydrogen. Inasmuch as 
the first part of the sputtering may be erratic and the dis- 
charge unsteady, it is well to cover the mirror with mica 
until sputtering has definitely started and become stable. 
This mica is mounted on pivots with an attached iron arma- 
ture, so that it can be operated with the help of a magnet 
through the walls of the bell jar; or it may be operated by 
tipping the whole system. 



The pressure for sputtering is usually adjusted so as to 
give a dark space of about the same length as the distance 
of the mirror from the cathode. 

The cathode should be shaped so that the boundary of 
the dark space is roughly parallel to the mirror surface. 
For flat or nearly flat mirrors the cathode is made flat, 
while for strongly curved mirrors it should be correspond- 
ingly curved. A U-shaped sheet cathode can be used for 
coating the two sides of a plate at once, and a central wire 
cathode can be used to coat the inside of tubes, provided 
that their length is not much greater than their diameter. 
Conversely, a cylindrical cathode can be used for coating 
fibers on all sides at once and for coating the outside of 

The gas admitted, when fast pumps are used, may be air, 
hydrogen, or argon. Hydrogen is preferred by some even 

Sputtering Rates of Metals 



Rate of Sputtering in Descending Order 



Pd, Au, Ag, Pb, Sn, Pt, Cu, Cd, Ni, Ir, Fe, 
Al, Mg. 


N 2 

Ag, Au, Pt, Pd, Cu, Ni. 



Cd, Ag, Pb, Au, Sb, Sn, Bi, Cu, Pt, Ni, Fe, 
W, Zn, Si, Al, Mg. 

Guntherschulze . . . . 

H 2 ° 

Bi 1470, Te 1200, As 1100, Tl 1080, Sb 890, 
Ag 740, Au 460, Pb 400, Zn 340, Cu 300, 
C 262, Sn 196, Fe 68, Ni 65, W 57, Co 56, 
Mo 56, Mn 38, Cd 32, Al 29, Cr 27, Ta 16, 


o 2 ° 

Zn 1030, Tl 650, Ag 614, Au 423, Pb 320, 
Cu 236, Sn 227, Fe 86, Mo 80, W 49, Ni 52, 
Cd 28. 

Crookes, Sir W., Roy. Soc, Proc, 50, 88 (1891). 

Kohlschtitter, V., Zeits. /. Elektrochem., 15, 316 (1909); Jahrb. Radioaktivitat, 
9, 335 (1912). 

Guntherschulze, A., Zeits. f. Physik, 86, 563 (1926), 38, 575 (1926). 

° Numbers give rate of sputtering in milligrams per ampere hour under 
conditions of cathode fall of 770 volts and current density of about 7 milli- 
amperes/cm 2 . 



though it has a very slow sputtering rate. The hydrogen 
may be obtained from a tank or from a gas electrolysis 
chamber. The relative sputtering rates for the various 
metals with different residual gases are given in Table I 
and Fig. 6. 

J y y /^*' — "AluiP* nurn 

O lOOO xooo 3000 400O 

50O lOOO 1500 

sputtering potential in volts 

Fig. 6. Guntherschulze's measurements of sputtering rates. 

E. O. Hulburt n has recently made a study of sputtering. 
He determined the rates of sputtering in a residual atmos- 
phere of air at a pressure giving 5 cm dark space. The 
voltage he used was 1000 to 3000 volts and the current 50 
milliamperes. The cathode was 5 cm in diameter and 2 to 
4 cm from the surface coated. His results are given in 
Table II. 

Time to Obtain Metal Films by Sputtering 



Sb, Bi, Cd, Au, Pb, Pt, Ag, Sn, Zn 

Co, Cu, Ir, Fe, Ni, Se, Te 

Opaque coating in 1 hour 
Opaque coating in 2 hours 

Mo, Ta, W 

Al, Be, C, Cr, Mg, Si 

Opaque coating in several hours 
Extremely low sputtering rate 

Hulburt states that the use of mercury vapor enormously 
increases the sputtering rate of chromium, aluminum, and 

11 Hulburt, E. O., Rev. Set. Instruments. 5. 85 (1934). 


silicon. Optical films of these metals were produced in less 
than 15 hours in this vapor. Good but not entirely opaque 
optical films of beryllium were obtained after sputtering for 
60 hours in hydrogen and mercury vapor. 

Clean dry surfaces and breath figures. To get a surface 
both clean and dry as required for sputtering and evapo- 
ration is a great deal more difficult than to clean it for chemi- 
cal silvering as described above. Most surfaces cleaned 
and then dried with absorbent cotton or a towel are found to 
condense the breath in a gray film. The reason is that in 
the drying process the glass surface becomes coated with a 
layer of contamination, which is probably a monomolecular 
film of fatty acid gathered from the cotton. Water con- 
denses on such a film in tiny droplets, while on a really clean 
surface it condenses in an invisible uniform film. 

Surfaces can be chemically cleaned and dried in a des- 
iccator. Such surfaces give a continuous deposit when 
breathed on. Also, surfaces may be dried with linen without 
contaminating them, as Wm. B. Hardy has succeeded in 
doing. Hardy found it necessary, however, to use linen 
from which the oily compounds had been extracted with 
pure benzene. 

However, a method to remove the contamination picked 
up from the towel when the mirror is dried is more practical 
than to depend upon successfully avoiding such contamina- 
tion. This dry cleaning can be effected by the action of ions. 

The study of this action of ions on the surface of glass 
started with Aitken and Lord Rayleigh. 12 They found that 
when the tip of a blowpipe flame was passed quickly over the 
surface of the glass, it cleaned the surface and produced a 
so-called breath figure; that is, if one breathed on the glass, 
the moisture condensed in a gray film of fine droplets, except 
that where the flame had traversed the surface, the moisture 
condensed in the form of a continuous "black" film. T. J. 
Baker and others have carried the study of breath figures 

12 Lord Rayleigh, Scientific Payers, Vol. 6, pages 26 and 127. Cambridge: 
University Press, 1920. Aitken, Roy. Soc. Edin., Proc, 94 (1893). 


further. 13 For example, Baker found that they were pro- 
duced only by the hotter flames, which are rich in ions. 
Among the interesting phenomena revealed by his investi- 
gation was that breath figures could also be produced by 
sparks, and that, curiously, they could be transferred from 
one glass plate to another if the two plates were held together 
but not quite in contact. He also discovered that the black 
area is a relatively good conductor of electricity and that the 
coefficient of friction between glass and glass was very high 

in the black area. Fig. 7 
"flamed" zone —glass point illustrates a simple experi- 
chatters and scratches ^ ment for demonstrating this 

difference in friction be- 
gi ass y/^^k}^ cine tween glass which has been 
P ,at ^/ F^l 3^ n ° flamed and that which has 

not been flamed. 

A. C. F. Pollard 14 found it 
easy to obtain good adherent 

"STlfd"" mo"ot?!^ S fil ™ °f chemical silver on 

without chattering glass by passing a blowpipe 

Fig. 7. over the surface of the glass 

before immersion in the sil- 
vering solution. He also found that for a short time a freshly 
fractured glass surface condenses moisture in a continuous 
black film. 

As a parallel to Pollard's discovery, it was found that alu- 
minum coats prepared by evaporation in vacuum adhere so 
tenaciously to areas that have been flamed that they cannot 
be removed by stripping Scotch tape off the film, although 
the tape removes the aluminum from regions not traversed 
by the flame. 15 Also, the black type of condensation, as 
well as good adhesion of an aluminum film, occurs after a 
glass surface is exposed to sparks at atmospheric pressure or 
to a glow discharge at reduced pressure. The explanation 

» Baker, T. J., Phil. Mag., U, 752 (1922). 

14 "The Making of Reflecting Surfaces," a discussion held by the Physical 
Society of London and the Optical Society, November 26, 1920. 

15 Strong, J., Rev. Sci. Instruments, 6, 97 (1935). 




grits and. 

of all these phenomena is that the ions of the hotter flames, 
sparks, or glow discharges clean the surface of the glass. 

The practices adopted to effect a final cleaning of a glass 
surface are either to expose it to the brush discharge from 
the electrode of a high-frequency transformer at atmos- 
pheric pressure or to expose the glass to a glow discharge in 
an evaporation chamber while it is being evacuated. 

Cleaning mirrors for aluminizing. When aluminum is 
deposited on a glass surface which is not adequately cleaned, 
che adhesion will be inferior 
to that exhibited by a coat on 
a properly cleaned surface. 
In most cases the mirror will 
look good at first but will 
develop countless tiny blisters 
after standing a day or so. 

The first phases of the clean- 
ing procedure for aluminizing 
are like those for chemical sil- 
vering. The preliminary 
cleaning with the rubber eraser 
is carried out with particular 
thoroughness. Small bubble holes in the face of the mirror 
that contain rouge and pitch from the figuring should be 
ground out with emery as shown in Fig. 8. If the rouge 
and pitch in small bubble holes is not removed, the towel 
used for drying the mirror may pick up some of the pitch and 
spread it over the surface of the mirror face in layers too 
thick to be removed by electrical cleaning. 

After the glass has been cleaned and rinsed as described 
above for silvering, it is dried with clean cotton towels. 
It is well to use old cotton towels, because after many 
launderings they become more absorbent and contain less 
fatty substances than absorbent cotton. Care is exercised 
to avoid contaminating the freshly laundered towel by 
touching it with the hands in the areas to be used to dry 
the mirror face. 

Fig. 8. 


Finally, the glass is exposed to a glow discharge during the 

Evaporation. The evaporation method for producing 
thin films on glass, quartz, and so forth, is simple both in its 
mechanism and in its practical application. A small piece 
of the metal (or nonmetal, for that matter) is simply heated 
in a high vacuum until its vapor pressure is about 10~ 2 mm 
of mercury or greater, whereupon it emits molecular rays in 
all directions. The degree of vacuum required for success- 
fully carrying out the process is such that the mean free 
path of the molecules is larger than the diameter of the 
vacuum container. Therefore molecular rays propagate 
from their source without disturbance until they impinge on 
the walls of the vacuum or some object within them. The 
mirror surface to be coated is exposed to these molecular 
rays, which condense on it to form the desired film. An 
interesting feature of the condensed film is that it apparently 
exhibits the same degree of polish as the underlying glass 
and so requires no subsequent burnishing, as does chemical 
silvering. Also, this film forms without material heating of 
the mirror. 

Although the evaporation method was known by 1912, it 
remained obscure, for some reason, long after it should have 
become a practical "tool" in the laboratory. 16 Among the 
items which have influenced its recent rather extensive 
applications are the development of a bare tungsten heater 
technique, 17 the adaptability of the process to nonmetals and 
for the application of aluminum, 18 and the development 
of high-speed vacuum pumps. (See Chapter III.) 

Whether or not a particular material is suited to giving 
films by the evaporation process is determined by the 
thermal stability and vapor pressure of the material and 
the practicality of bringing the material to the evaporation 
temperature in vacuum. 

16 Pringsheim, P., and Pohl, R., Deutsch. Phys. GeselL, Verh., U, 506 (1912). 

17 Ritschl, R., Zeits.f. Physik, 69, 578 (1931). 

18 Strong, J., Astrophys. J., 83, 401 (1936). 



Tungsten heaters useful for bringing some of the metals 

to the evaporation temperature are shown in Figs. 12 and 

15 to 20. The evaporation temperatures of the metals are 

given in Table III. 


Evaporation Temperature ° of Different Metals 


Evaporation Temperature 
T° Absolute 


Evaporation Temperature 
T° Absolute" 























































a Temperature at which vapor pressure equals 10 -2 mm of mercury. 

Baur, E., and Brunner, R., Helv. chim. Acta., 17, 959 (1934). 

Espe, W., and Knoll, M., Werkstoffkunde der Hochvakuumtechnik, page 358. 
Berlin: Julius Springer, 1936. 

Knoll, M., Ollendorff, F., and Rompe, E., Gasentladungs-Tabellen. Berlin: 
Julius Springer, 1935. 

Landolt-Bornstein, Phys. Chem. Tabellen, Fifth Edition. Berlin: Julius 
Springer, 1923-1936. 

Leitgebel, W., Metallwirtschaft, U, 267 (1935). 

Most of the metals melt first before they evaporate, the 
molten metal being kept from falling out of the coil by sur- 
face tension. 

Other metals, like magnesium, sublime. Of these, some 
sublime very slowly, because the metal will not fuse to the 
tungsten wire in vacuum. Chromium affords an example. 
The evaporation of such a metal is managed as follows: 
It is first brought to fusion temperature in the tungsten 
coil in an atmosphere of hydrogen or helium. These gases 



facilitate heat transfer between tungsten and the chromium 
or other metal, and, in addition, they restrain evaporation 
of the metal. (See Fig. 9.) After intimate contact with the 
tungsten wire is established, the metal will then sublime 

copper tube 

source of 

coil to be"pre-f usecT^ 

6 copper-wire supports 
to power source soldered to copper tubes 

Fig. 9. Arrangement for pre-fusion of metal to tungsten coil. 

faster in the vacuum, because the heat is transmitted to it 
more effectively. An alternate way of attaining the same 
end is to electroplate the chromium or other metal onto the 
tungsten coil. 19 The metals best managed by the above 
procedures include, besides chromium, the platinum metals 
and beryllium. 

Frequently, it is desirable to prefuse a metal which other- 
wise sublimes, in order to free it from included impurities. 

Such metals as calcium, mag- 
nesium, and cadmium can be 
prefused in helium to outgas 
them and to prepare them 
for evaporation. 

A great many metals react 

with the tungsten coil, as, for 
soft copper . ° • i i u it 

"supporting - wire example, iron, nickel, berylli- 

Fig. 10. um, chromium, the platinum 

19 This electroplating technique is apparently one which has been frequently- 
used. Note the following references on its application to platinum and 
chromium respectively: 

Strong, J., Phys. Rev., 89, 1012 (1932). 

Williams, Robley C, Phys. Rev., 41, 255 (1932). 


metals, and aluminum. In spite of this, it is possible to 
evaporate them for the preparation of small laboratory 

Fig. 10 shows a neat simple insulated support for wires 
in vacuum. 

Evaporation technique for aluminum. The technique 
for evaporation of aluminum from tungsten coils is of special 
interest, since this metal is important for surfacing where 
high ultraviolet and high visible reflectivity are desired in 
combination with freedom from tarnishing. 

Pringsheim and Pohl discovered that several metals 
(including aluminum) could be evaporated in vacuum and 
condensed on a glass surface to form a polished reflecting 
film. They used a magnesia crucible from which to distill 
the metal. 20 R. Ritschl, in 1928, in making an application of 
the evaporation method to the preparation of half-silvered 
interferometer mirrors, heated the silver in a bare tungsten 
coil. 21 This change in technique has the advantage that the 
tungsten does not evaporate or outgas so much in a vacuum 
as does the magnesia crucible. 

Following this, Cartwright and Strong developed a simple 
apparatus for carrying out the evaporation process in the 
laboratory and made a survey of its applicability to different 
metals. 22 The usual technique, in which the metal to be 
evaporated was heated in a helix of tungsten wire, was found 
successful, except with the metals aluminum and beryllium, 
which dissolved the tungsten coil. 

Other attempts were made to develop this technique of 
evaporating aluminum. 23 Experiments were carried out 
with crucibles of graphite, pure fused magnesia, and alumina 
(sapphire), as well as with sintered and fused crucibles of 
thorium oxide. These experiments showed that heating in a 
crucible was apparently impractical, since either the metal 

20 See footnote 16. 

21 See footnote 17. 

22 Cartwright, C. Hawley, and Strong, J., Rev. Set. Instruments, $. 189 

23 Cartwright, C. Hawley, Rev. Sci. Instruments, 3, 302 (1932). 



reacted chemically with the material of the crucible or the 
latter evaporated when the aluminum was heated. 

The discovery that tung- 

mandrel - £j/i6 round 

4" long 


groove -'A" pi tch /* 


sten has a limited solubility 
in molten aluminum led to 
the bare tungsten method of 
evaporation — the most prac- 
ticed of all the methods. 24 

A chemical analysis of the 
tungsten alloy that is 
formed when aluminum is 
fused on a tungsten coil 
showed the solubility of 
tungsten in aluminum to be 
about 3 per cent by volume. 
Accordingly, the burning out 
of the tungsten wire may be 
avoided by the simple expe- 
dient of making it of rela- 
tively large diameter and ar- 
ranging the charge so that the solubility of the molten alumi- 
num for tungsten can be satisfied without dangerously reduc- 
ing the diameter of the wire. 

It might be expected that 
some of the dissolved tungsten 
would boil away, especially 
since its spectrum has been 
observed during evaporation. 25 
In order to test this point, 
a coil was weighed before 
and after evaporating several 
charges of aluminum. Instead 
of a loss in weight, an in- 
crease was observed, indicat- 
ing that some aluminum had 

Fig. n. 

coil after preliminary firing 

i i l ? 

scale of inches 
Fig. 12. 

24 Strong, J., Phys. Rev., 43, 498 (1933). 

25 Gaviola, E., and Strong, J., Phys. Rev., 48, 136 (1935). 



diffused into the tungsten. However, extended heating in 
vacuum at a very high temperature decreased the weight, 
until, within the experimental error, it became the same as 
in the beginning. A chemical analysis of the condensed 
metal film was made to test whether or not tungsten is 
evaporated. The analysis gave no definite indication of 
tungsten. A concentration of 0.03 per cent by weight was 

to power ± - 

inner shell 
of brass to 
carry switch 
and filaments 


handle for operating 
switch and movable 
baffle ^ 

Shackle for 
lifting bell jar 



steel bell 


for lead 
fuse wire 

steel base plate with 
reinforcing ribs 

° scale of feet t 

Fig. 13. 

detectable. The tungsten which is dissolved thus appears to 
be almost completely precipitated back onto the coil as the 
evaporation proceeds. Although it may not be deposited 
back in exactly the same place, it does compensate in a large 
measure for the decrease in diameter of the tungsten wire. 
The arrangement used at first for aluminizing mirrors at 
the California Institute of Technology is shown in Figs. 11 
and 12. It is in the form of a helix, consisting of 10 turns of 
30-mil tungsten wire., ^ of an inch in diameter and pitched 



4 turns to the inch. A U-shaped piece of aluminum wire 
1 mm in diameter and about 10 mm in total length is 
clamped to each turn as is shown in Fig. 11. A potential 
of 20 volts applied to the coil in vacuum for 4 seconds prefuses 
these pieces as shown in Fig. 12. At this stage, surface 
tension keeps the molten aluminum from dropping. This 
prefusion also serves to free the metal from oxide and other 
impurities. It is customary to make a separate run in 
order to effect this fusing of the aluminum to the tungsten 

Fig. 14. 

wires. In the 40-inch tank (see Fig. 13), however, the coils 
are covered by a baffle during the preliminary firing. The 
aluminum is finally distilled from the coils by applying the 
same voltage to each coil for about 15 seconds. 

Actually, the aluminum does not evaporate from the 
fused metal but from the adjacent tungsten wire. This is 
clearly shown by the "self -photograph" of the filament re- 
produced in Fig. 14. This "self -photograph' 7 was recorded 
on glass with the molecular rays of aluminum passing 
through a pinhole. 

A recently developed evaporation source allows a much 



3 strands 
of 20-mtl 

higher rate of evaporation of aluminum with less tendency 

to burn out or drop molten aluminum. The new source 

uses three or four 20-mil tungsten wires twisted together as 

shown in Fig. 15. The metal 

charge, applied as illustrated in 

Fig. 11, flows out to fill the space 

between the wires when heat is 

applied. The aluminum covers 

the tungsten completely, so that 

a minimum "ratio' ' of heat radi- 
ation to molecular radiation of 

aluminum is achieved. 

Fig. 16 shows the form by which 

the new source is applied to the 

evaporation of gold. When the 

gold melts in the "cup," it is 

drawn out to coat the tungsten and 

it fills up the spaces between wires 

from one end to the other. 

For evaporation of silver and 

copper the source should be 

made from tantalum or molyb- Fig. 15. 

denum rather than tungsten, as 

the latter metal is not easily wet with silver and copper. 
For evaporation of the platinum metals, a unit similar to 

the one shown in Fig. 15 is made up of three 20-mil tungsten 

wires and one platinum metal wire of the same diameter. 

The "ratio" of heat to metal 
radiated is a minimum. 
Furthermore, the awkward 
process of electroplating the 
platinum on the filament is 
avoided. The evaporation 
should proceed slowly, even 

from this source, because if too much current is applied, the 

evaporation is no longer smooth, and globules of metal are 

discharged from the source. 

scale of inches 
Fig. 16. 



Chromium is easily evaporated from a source like the one 
shown in Fig. 16. A piece of the metal is put in the "cup" 
and is preheated in an atmosphere of hydro- 
gen or helium to fuse it and distribute it over 
the tungsten. Various other evaporation 
sources are illustrated in Figs. 17 to 20. 
Vacuum equipment. The evaporation 
process is carried out in a vacuum of 10 -3 
mm of mercury or 
better. For small 
mirrors the neces- 
sary vacuum may be obtained with a 
kinetic pumping system such as the 
one shown in the previous chapter. 

The 40-inch tank, Fig. 13, shows the type of equipment used 
at the California Institute of Technology for larger mirrors. 

Still larger systems have 

T 13^ turns -20- 

rnil wire - 
\ % mandrel 

metal to be 
inserted in coil 

Fig. 17. 

metal to 
be evapo- 
placed in coil 

Fig. 18. 

flat tungsten strip with dent 
for metal to be evaporated 

t3> s .S* 





flat tungsten folded as 
shown to form a crucible 
for metal to be evaporated 

Fig. 19. 

been used. 26 

The cleaning electrode 
shown in Fig. 13 allows the 
vacuum vessel, containing 
the mirror, to be filled with 
a glow discharge during the 
preliminary evacuation with 
the roughing pumps ; and this 
discharge effects the final 
cleaning of the mirror face. 

It is recommended that the 
aluminum be evaporated 
soon after a nonconducting 
vacuum has been reached, in 

26 Strong, J., Astrophys. J., 83, 401 (1936). 

Metal tanks of seamless steel are available from the Eclipse Fuel Engineering 
Company (Los Angeles agent, James H. Knopf) in the same form as bell jars. 
After the foot is machined, they are suitable for sealing to a base plate to form 
a good vacuum container for evaporation. It is advisable to clean the tank 
inside and out by sand blasting and to coat it inside with Apiezon wax "W" 
and outside with Glyptal lacquer. 


order to obtain maximum tenacity between the aluminum 
film and the glass. Also, this procedure yields harder films. 

Uniform films. In order to obtain a uniform coat on 
large mirrors, aluminum is evaporated from several tungsten 
sources suitably arranged, rather than from one movable 

The evaporation of polonium in a high vacuum from a 
point source has been investigated by Bonet-Maury. 27 
This metal was chosen on account of its radioactivity. He 
found that the condensation on a plane sur- 
face is proportional to the inverse square of |S|f] (j|| 
the distance from the source, and to the L ~j| \^\ 

cosine of the angle between the normal to J I j W^ 

the surface and the line connecting the sur- * \^ ^5p 
face with the source. We may assume that tantrum cap to 
the same is true of other metals which have bS C u?ed Tor'e!**^ 

i j. j. j. orating oxides or 

a low vapor pressure at room temperature. the dent in the 

Starting with this assumption, we may rnet&i? ° r 
consider the distribution of the film thick- Fi 20 

ness r produced by various experimental 
arrangements. In the case of evaporation to the inside 
surface of a sphere of radius p from a point source of vapor 
at its center, the situation is ^very simple. We get a uni- 
form film of which the thickness r is 

T0 = 4^' (1) 

Here m is the mass of metal evaporated and d is its density. 
The film thickness at P on a plane surface at the normal 
distance p from a point source of evaporation is 

■>-sb «■'-«<?)'' » 

Here r is the thickness at P, r is the distance from the 
source to P, and 6 is the inclination of the surface P to the 
molecular rays emitted by the source which impinge on it 

27 Bonet-Maury, P., Ann. de Physique, 11, 253 (1929). 



The film thickness produced on a plane surface by a 
circular array of vapor sources can be determined by apply- 
ing the above formula to each of the sources. (See Fig. 21.) 
If there are N coils spaced uniformly around a circle at a 
distance p from the surface to be coated, the film thickness 

on the surface at P, which is 

taKen as unity 

at a distance a from the in- 
tersection of the axis of the 
circle with the face of the mir- 
ror, is given by the expression 

T P = 

M P 

N 1 

2u js 


4xdN f r\ 

Here M is the total mass of 
metal evaporated, and r is 
the distance from P to the 
coil represented by the sum- 
mation index i. 

Dr. Edward M. Thorndike 
made the same calculation, 

assuming a continuous circular source. The thickness is 

given in this case by 

Fig. 21. 

T P = 


87T 2 d 



2 *dd 


Here the point source at distance r from the point P is 
replaced by a line source represented by the angle element 
d6 at distance r, as before. This calculation involves the 

Jo r 3 Jo ( 


o (1 f a? + p 2 

2a cos 0) 3/2 


[(a - l) 2 + p*-]-V(a + l) 2 + p 2 \V(a + l) 2 + p\ 


in which E represents the elliptic function. 28 Values of this 
integral calculated by Thorndike are given in Table IV. 

28 Bierens de Haan, David, Nouvelles tables d'integrales definies, Table 67, 
Eq. 3, page 102. Leyden: P. Engels, 1867. 



Values of 

o r* 


fob Various Parameters 



P = 1 

p = 1.1 

p = 1.2 

P =2 






























































For convenience, the radius of the circular source is here 
taken as unity. We see from this table that f or p = 1 the 
film is quite uniform as far out from the center as a = 1. 
This case was realized in the 40-inch aluminizing tank by a 
circular array of twelve of the standard coils (see Fig. 12) 
spaced around a circle 36 inches in diameter, 18 inches above 
the face of the astronomical reflector to be coated (Fig. 22). 
Tests of transmission of a film produced with partially loaded 
coils confirmed the calculation, since the coat exhibited the 
expected uniformity. 

12 filaments 


24 filaments 
4 „U filaments 
\ T >»4 filaments 

Fig. 22. Arrangements of evaporation coils for large mirrors. 


In a larger 108-inch tank it was not convenient to use a 
similar array of coils spaced 50 inches from the face of the 
mirror. Instead, three arrays were used, each about 20 
inches from the mirror. The arrangement is shown in 
Fig. 22. From the expressions developed above, as well 
as from actual tests, it was found that four coils in the 
center, twelve on a circle of 50 inches in diameter, and 
twenty-four on a circle of 100 inches in diameter gave the 
proper loading. This arrangement produced a uniform 
film of proper thickness on a 100-inch mirror, the film being 
just a little thicker than that required to be opaque to sun- 
light. It is desirable to have this thickness (about 1000 A), 
since much thicker films are more easily scratched, while 
thinner ones may in time become transparent as a result of 
the gradual growth of thickness of the oxide layer which 
forms on the aluminum coat. 

Parabolizing a spherical mirror with aluminum. As soon 
as the technique for the attainment of uniform films was 
perfected, it became possible to prepare nonuniform films, 
with the thickness of the film varying in just the manner 
required to parabolize a spherical mirror. The difference 
r between the circle and parabola illustrated in Fig. 23 is 
given to close approximation by the expression 

r = y\yl - y*)^f 3 > (6) 

where y is the ordinate and R is the radius of curvature of 
the circle. y represents the ordinate where the two curves 
intersect. The difference is zero at y — and at y = y Q 
and has a maximum at y = y /\ / 2. 

If a spherical mirror of diameter 2y (represented by the 
surface generated by rotation of the circle in Fig. 23 about 
the X axis) is to be transformed to a paraboloidal surface 
(the surface generated by rotation of the parabola), it is 
evident from Eq. 6 that it is necessary to add to the sphere 
a zone of aluminum which has its maximum thickness at 



y =- y / V2, tapering off on either side of this as required by 
the equation. 

The maximum thickness of aluminum, T max ., required de- 
pends naturally upon the radius of curvature of the 


Fig. 23. 

sphere, R. The connection between r r 
given by the expression 

or, in terms of its / value, 

32# s 

2048/ 3 ' 

R, and y is 



Inasmuch as it is possible to put down films of aluminum 
to lfj, thickness and greater, it is possible to parabolize a 
12-inch mirror //6, which requires a maximum thickness of 
only 0.34/x of aluminum. This is not an uncommon ex- 
ample encountered in astronomical mirrors. 



The correct procedure for applying such a parabolizing 
film is first to compute the thickness and distribution of the 
aluminum film produced by a point source positioned op- 
posite the center of the mirror as shown in Fig. 24. This 

com m utator 

ball bearings 

sof t-iron 

mirror to be 



to rotate 

brass plate 

to carry mirror 

baffle to 


film thicKness 



to vacuum pumps 

Fig. 24. 

can be done by the use of the formula given below for the 
thickness of aluminum r produced at a distance y from the 
center of the mirror. 




Here m is the total mass of aluminum evaporated, in grams, 
and d is the distance between the source and the point in 
question on the mirror face. 

A baffle of the shape illustrated by Fig. 25 is then cut from 
thin sheet brass and placed directly in front of the mirror as 
shown in Fig. 24. This baffle can be rotated, or, what is 



Fig. 25. 

more convenient, it may be fixed and the mirror rotated as 
shown in Fig. 24. The baffle is so designed as to modify 
the thickness which would 
otherwise be obtained (given 
by Eq. 9), so that it will con- 
form with that required by 
Eq. 6. The baffle will have 
zero angular opening at the 
center and edge and a maxi- 
mum opening very near to 
y = 2/o/V2. It is to be re- 
membered that the effect of 
the baffle in a given zone is 
to decrease the thickness by 
a factor which is the ratio of 
the quantities, 360° minus the angular opening of the baffle 
opposite the particular zone in question, to 360°. In order 
to avoid astigmatism, the mirror is rotated a great many 
times during the deposition. 

It is necessary, for some reason not yet clearly demon- 
strated, to evaporate slightly more aluminum than the simple 

theory outlined above pre- 
dicts. The procedure in this 
case is to deposit some metal 
(about the theoretical amount) 
and then test the mirror. On 
the basis of the Foucault test, 
an additional amount is 
evaporated, and so on until 
the required figure is obtained. 
If too much metal is added, 
the coat can be washed off 
with caustic soda. Usually 
the mirror can be finished 
on the second attempt. 
When several mirrors, all alike, are to be parabolized, this 
preliminary testing may be done once for all. 

Fig. 26. Starting sphere tested at 
the center of curvature. 



Fig. 27. 

Sphere tested at its mean 

Figs. 26, 27, and 28 show f ocograms of a mirror parabolized 
by this method. It was originally a sphere true to Jq- of a 

wave length of green light, as 
the first focogram (Fig. 26), 
taken at its center of curva- 
ture, shows. This sphere was 
152 j inches in radius of cur- 
vature. 2y was 12| inches. 
The next focogram, Fig. 27, 
shows it at its mean focus 
when tested with parallel 
light with the aid of a testing 
flat, obviously in need of para- 
bolizing to give a good knife- 
edge cutoff. After it was para- 
bolized with a coat of alumi- 
num, it appeared as shown in the third focogram, Fig. 28. 
Here, again, it exhibits a true figure of revolution, this time 
a parabola true to less than ^ of a wave length of green light. 

Mirrors imperfectly figured 
by conventional methods 
can be improved by this pro- 
cedure. In this case the baffle 
design is determined by a pre- 
liminary quantitative survey 
of the mirror with a knife- 
edge testing outfit. (See 
Chapter II.) 

It is possible to apply a thin 
film of aluminum to a convex 
sphere and transform it to a 
hyperbolic figure of revolution 
for use as the secondary mir- 
ror in a Cassegrain telescope. 

The formula for the difference between the hyperbola, or 
any conic of eccentricity e, and the sphere tangent to it at 
the center and touching it at the radius distance y is 

Fig. 28. Sphere after paraboliz- 
ing with an aluminum film. Tested 
at the focus. 



= fyKvo - y 2 ) 



Eq. 6 for the parabola is Eq. 10 when e = 1. To obtain a 
hyperbola, it is necessary to have the aluminum thick at 
center and edge with a minimum at y = y /^/2. The baffle 
to effect this is just the inverse of the one shown in Fig. 25, 
being open where the other is opaque and vice versa. The 
further details of the process are described in a paper by 
Strong and Gaviola and in the paper of Gaviola on the 
quantitative use of the knife-edge test. 29 

Partially reflecting films. Partially reflecting films of 
silver and aluminum are useful for dividing a beam of light 
in many optical instruments such as color cameras and inter- 

Figs. 29 and 30 show the reflection and transmission char- 
acteristics of silver and aluminum films obtained by the 

20 10 SO 80 lOO 120 140 160 180 200 

mg of Silver at £7 cm 
Fig. 29. 

evaporation of various amounts of metal. The curves 
illustrate the color characteristics of the films and their 
efficiencies. They also indicate approximately the amount 
of metal to be evaporated to obtain any desired ratio of 
reflection to transmission. The curves for silver refer to 

29 Strong, J., and Gaviola, E., J.O.S.A. 
Gaviola, E., J.O.S.A., 26, 163 (1936). 

153 (1936). 



fresh deposits, whereas the curves for aluminum apply to 
films about 6 months old, which have more or less attained 
their equilibrium optical characteristics. 

The reproducibility with which any given film can be 
prepared from the information given in Figs. 29 and 30 is 


8 lO 12 14 16 

rhg of Aluminum at 33 cm 
Fig. 30. 



unfortunately not very great. The variations to be expected 
are greater in the case of aluminum. 

The films from which the curves in Figs. 29 and 30 were 
obtained were evaporated with a vacuum of 1 to 5 X 
10~ 5 mm, the mirror distance being 33 cm in the case of 
aluminum and 27 cm in the case of silver. A source like the 

one shown in Fig. 17 was used 
for silver. The metal was in 
the form of a 40-mil wire. 
A straight, horizontal 30- 
mil tungsten wire served as 
the evaporation source for 
aluminum as shown in Fig. 31. The metal was a weighed 
U-shaped piece of wire pinched onto the center of the 
tungsten wire. 

Silver films have a greater efficiency than aluminum films, 
and they are, accordingly, best for coating Farby and Perot 
interferometer plates. They may be protected from the 

'"*>* heavy copper wires*** 
Fig. 31. 


tarnishing gases in the atmosphere by a thin layer of cal- 
cium fluoride or quartz. 

The calcium fluoride (or quartz) films should be about 
| of a fringe in thickness. If a copper sheet is placed close 
to the evaporation source, it is possible to count the fringes 
as they are formed on this sheet by the evaporated calcium 
fluoride (or quartz). The square of the ratio of the distance 
of the copper to that of the silver gives the ratio of film 
thickness of calcium fluoride (or quartz) evaporated onto 
these two surfaces. The evaporation of calcium fluoride (or 
quartz) is stopped after an appropriate number of fringes 
have appeared on the copper. 

A thin film of aluminum on the silver will oxidize to a pro- 
tecting layer of aluminum oxide on exposure to the air. 
The proper amount of aluminum to be evaporated is about 
one-sixteenth the amount required to give a half -transmitting 
coat. Accordingly, the proper amount of aluminum may be 
gauged by means of an auxiliary glass plate positioned at 
one-fourth the silver film distance from the evaporation 
source. The proper amount of aluminum is evaporated 
when the film on the auxiliary glass plate appears to be about 
half -transmitting. 

When a half-silvered mirror on glass is cemented with 
balsam to a second glass surface, the ratio of transmission 
to reflection is increased by about 5 per cent. 


The Use of Fused Silica 


H. V. Neher 

General remarks on fused quartz. Formerly made only 
in rod and tube form, fused quartz 1 is now often employed 
as a substitute for glass in chemical ware, and most of the 
common pieces used in chemistry are now obtainable in 
this material. Such articles as flasks, beakers, dishes, 
plates, and so forth, are in fairly common use. 

Apparatus made from fused quartz has two chief ad- 
vantages over that made from glass. The low thermal 
expansion coefficient eliminates all fear of breakage due to 
rapid temperature changes. A hot piece of quartz plunged 
into water suffers no ill effects. Also, its relatively high 
melting point makes possible the study of reactions which 
would be more difficult with glass. 

As will be pointed out later, many of its properties make 
it valuable in instruments of various kinds and when con- 
stancy is a prime requisite. One particularly valuable prop- 
erty of fused quartz is its extremely low loss of energy due 
to internal friction when stresses are applied. The loss 
amounts to only 10 -3 of that in the best of the metals. 
Another property of value lies in its constancy of length. 
It not only has an extremely small thermal expansion 
coefficient, but returns to its original length after having 
been heated or cooled. 

The chief disadvantage of fused quartz is its high cost, 

1 Fused quartz is obtainable from the Thermal Syndicate and the General 
Electric Company. Each carries a large stock of quartz products and will 
make special pieces on demand. 



due mainly to the fact that it has a high melting point and 
demands special methods for its manufacture. The fact 
that it fuses with difficulty makes the working of tubing 
greater than an inch in diameter more or less impractical in 
the laboratory. Although an oxyhydrogen flame becomes 
useful when working large pieces of quartz, small pieces can 
be worked easily with an oxy-natural gas flame. An ordi- 
nary Bunsen burner flame using natural gas is hot enough to 
soften small pieces. 

A very useful property discovered by C. V. Boys 2 in 1889, 
and discussed in detail later, is that fused quartz can be 
drawn into fine fibers which have remarkable strength. 
Fibers of any size down to 1/x (0.0001 cm) diameter or less 
can be easily and rapidly produced. No other vitreous ma- 
terial can in any way approach fused quartz in performance 
when made into these fine fibers. 

Chemical properties. Fused silica at room temperature 
is inactive to practically all chemicals except hydrofluoric 
acid and the alkalies. However, at high temperatures it 
reacts with most metallic salts, forming silicates. This is 
due to the fact that silicon dioxide is an acid in the general 
sense of the term, and as such reacts vigorously at high tem- 
peratures with metallic oxides which are bases. The noble 
metals do not form silicates, and a quartz fiber covered with 
gold may be heated until the gold evaporates, without 
harming the fiber. 

Physical properties. Thermal properties. The coefficient 
of thermal expansion of fused quartz rod under no stress has 
been measured with considerable accuracy. 3 The mean 
values near room temperature, defined by a = {1/1) (AZ/fe — h) 
are given in Table I. For comparison, steel has a coefficient 
of 10.1 X 10~ 6 °C. _1 , or 25 times as large, while for Invar a is 
about 0.9 X 10 -6 °C. -1 . The coefficient of thermal expan- 
sion has not been measured for various sizes of fibers under 
varying amounts of strain. 

2 Boys, C. V., Roy. Soc, Phil. Trans., 148, 159 (1889). 
3 Kaye, G. W. C, Phil. Mag., 20, 718 (1910). 



The Mean Values of the Coefficient of Ther- 
mal Expansion of Fused Quartz Near Room 
Temperature, Defined by a =(l/l)(Al/i2 — h) 



(X lO-^C." 1 ) 

- 40 to 
Oto 30 
30 to 100 
100 to 150 


The coefficient of thermal hysteresis of fused quartz 
is less than for any other known material. If a substance 
of length I is heated from a temperature h to a tempera- 
ture U and allowed to cool to t h then (l/l)(Al/t2 — ti), where 
AZ is the residual difference in length, is a measure of the 
thermal hysteresis. For quartz, this quantity is —1 to 
— 5 X 10 _9 °C. -1 ; that is, it contracts more than it expands. 
In comparison, Invar has a similar coefficient of —100 X 
10 _9O C. -1 . This property makes fused quartz particularly 
valuable when it is necessary to maintain dimensions 
accurately. 4 

If fused quartz is held at a temperature above 1200°C. for 
some time, crystallization gradually takes place, beginning 
at the surface and working inward. As the temperature is 
raised, the crystallization becomes more rapid until a tem- 
perature is reached at which the crystals melt. When quartz 
is worked locally in a flame, a milky surface will form be- 
tween the soft quartz and the cool portion. This is probably 
due to condensation of evaporated quartz and does no harm 
to the material except in appearance. 

Elastic properties. The normal coefficient of elasticity, 
or the reciprocal of Young's modulus for quartz rod at room 

4 For a discussion of the behavior of metals and quartz used as standards of 
length the reader is referred to Glazebrook, Sir Richard Tetley, editor, Diction- 
ary of Applied Physics, Volume III, pages 471-475. New York: The Mac- 
millan Company, 1922-1923. 

Chap. VI 



temperature, was measured first by Boys. This coefficient 

is defined by 

A = * Al 
Y I ASn 

where Y is Young's modulus and S n is the normal stress. 
Boys found the value Y = 5.2 X 10 11 dynes cm~ 2 , which is 
very near the most recently determined values for fibers 
from 50/jl to 100/jl in diameter. Young's modulus varies 
with the size of the fiber, becoming greater as the size of 
the fiber diminishes. This variation can be expressed by 

Y = * 1U + 5.9 X 10 11 dynes cm" 2 , 

where d is the diameter of the fiber in microns. This rela- 
tion fails to hold, giving values too large, for fibers less 
than lOjii in diameter. Experimental values of Y for various 
sizes of fibers are given in Table II. The increase in modulus 
of elasticity with decrease in size is due to the importance of 
the surface layer for the smaller fibers, which has a different 

elastic constant. 


Breaking Strength, Young's Modulus, Y, Modulus of Rigidity, Z, 
and Al/l for Failure for Different Sizes of Quartz Fibers 



(X 10 11 ) 


(X 10 11 ) 


(X 10 11 ) 



for Failure 













































Data taken from Reinkober, O., Phys. Zeits., 38, 112 (1937). 

These are mean values; values of individual fibers may be as much 
20 per cent higher or lower than those given. Units are in dynes cm -2 . 


The tangential coefficient of elasticity, or the reciprocal of 

the rigidity modulus, for solid rod of radius r and length Z, is 

denned as 

I =, I A W 
Z 21 AS t ' 

where S t is the tangential stress and <£ is the angle of twist 
of the rod. For a uniform solid round rod S t = (L/r)/{irr 2 ), 
where L is the applied torque and r is the radius. Z has a 
minimum value of 3 X 10 11 dynes cm -2 but depends, as does 
Y, on the size of the fiber, as shown in Table II. 

Two other elastic quantities are very often useful. The 
first indicates how much a fiber can be stretched before it 
breaks, that is, 

(t) =¥• 

\ l /for failure * 

where (S H )< is the normal stress for failure. Values of 
AZ/Z for failure are given in Table II. These apply only 
to fresh, clean fibers or those which have been kept perfectly 
clean and dry. (See below as to how to preserve fibers.) 
As far as is known, no other material approaches this factor. 
For the best nickel-vanadium steels the ratio is about 0.01. 
A comparison of Young's modulus for each material shows 
that quartz fiber compares favorably in strength with the 
strongest materials known. 

The second quantity indicates how much a fiber can be 
twisted without failure, that is, 


'4) - fco.05 

*> /for failure ^i 

for fibers up to 20/z in diameter, where (&)/ is the tangential 
stress for failure. This ratio also increases as the size of the 
fiber decreases. Thus, a fiber 5/x in diameter can be twisted 
through at least 20 revolutions per centimeter of length 
before it fails. It should be remarked that the elastic limits 
for both normal and tangential stresses are coincident with 
the point of failure. 


Another property of quartz which enhances its value for 
electrometer and other suspensions is its low internal vis- 
cosity. If a fiber is twisted through an angle <£, then the 
shearing stress is not strictly a constant but depends on 
time, thus: 

Si = Z 2l+2dtVT)' 

The coefficient rj is a measure of the internal friction, or 
viscosity. Some representative values 5 are given in 
Table III. 


Viscosity of Various Solids 


(X 10 9 poises) 




Nickel .... 








0.001 (approx.) 

If a fiber of length I and radius r is allowed to oscillate 
in a vacuum with a body of moment of inertia I suspended 
from the lower end, and if T is the period and X the loga- 
rithmic decrement of the vibration, the coefficient of viscosity 
in poises is given by 

v Trr*T' 

If such a torsion pendulum has a period of 2 seconds, it will 
lose about 10 per cent of its amplitude in 24 hours. 

Thus rj, as defined above, should be as small as possible if 
the internal losses are to be kept at a minimum. 

5 Honda, K., Phil. Mag., 42, 115 (1921). 
Iida, K., Bull. Earthquake Res Inst, of Tokyo University, 13, 665 (1935) . 


Thermal-elastic properties. Both Young's modulus and 
the rigidity modulus for fused quartz depend on tempera- 
ture. Each becomes greater with moderate increase in 
temperature. Boys 6 gives the coefficient of Y as 1.3 X 
10 _4 °C.~ 1 , and for Z it is the same. For very accurate work 
any instrument using quartz fiber should be calibrated at 
more than one temperature. 

Hardness. Fused silica has a hardness of 7 on the 1 to 10 
scale. It is thus harder than glass and also harder than 
most of the metals. 

Surface tension of molten silica. If a fiber is heated until 
the quartz becomes quite soft, it will tend either to shrink 
and enlarge at the point of heating or to pull apart, depend- 
ing on the tension. We may define the surface tension as the 
force per unit of circumference tending to pull the fiber to- 
gether. This varies with the temperature, but an average 
value will be 250 dynes cm -1 . In comparison, glass has a 
surface tension of 140 to 160 dynes cm -1 . 

Electrical properties. When fused quartz is clean and dry, 
it is probably the best electrical insulator known. For 
this reason it is useful in such apparatus as electroscopes and 
electrometers, in which leakage must be reduced to a mini- 
mum. If used in the open air, quartz covered with the wax 
known as ceresin is still better than amber as an insulator. 
Care should be taken that the ceresin is that distilled from 
the natural mineral and not the synthetic material very often 
sold. When it is applied, the temperature of both the quartz 
and the ceresin should be from 80° to 100°C. for the first 
dip. Thicker coatings can be applied by allowing the 
quartz to cool before dipping again. 

The absorption of electrical charge, or "soak-in," is ex- 
tremely low, being less than 10 per cent of that for amber. 

The use of quartz in the form of fibers. The remarkable 
property of retaining and even increasing its strength as it is 
drawn into fine fibers makes the number of applications of 

6 Glazebrook, Sir Richard Tetley, editor, Dictionary of Applied Physics, 
Volume III, page 699. New York: The Macmillan Company, 1922-1923. 

Chap. V] 



quartz to fine instruments many and varied. Few scientists, 
it seems, have realized and appreciated its values. Stronger 
than any of the metals used for suspensions, with the ex- 
ception of tungsten, it has the advantage that it can be made 
according to the specific requirements. Although some 
practice is necessary to acquire the proper skill, its acquisi- 
tion would seem eminently worth while, considering the 
results that can be obtained. 

Equipment useful in making and working with quartz 
fibers. A description of the torch burning natural gas and 
oxygen used by the author of this chapter will be given. If 
other gases are used, it may be necessary to modify the 
technique given below to meet the specific conditions. 

The torch is made from a piece of brass tubing bent into 
the shape shown in Fig. 1 and having one end threaded for 

tips- %4 diam. ^fc'lonrf 
holes- .05 to o.Z mm 

torch for use in micromanipulator 

ty& brass >-. 

brass tubing 
'/te insi de d 

y4" tubing *»« fiber or Bake lite/ 

small torch for hand use 

large torch 

brass handle 

for blowing out fibers, etc. 

° scale of inches *° 

Fig. 1. Large and small torches for working fused quartz. 

removable tips. The best size of opening for quartz work 
is about 2 mm in diameter. Other sizes of tips from 1 to 3 
mm in diameter will be found useful. To produce the neces- 
sary long steady flame, the length of the hole in the tip 
should be at least five times its diameter. The oxygen and 
gas are mixed at some distance from the torch. An ordinary 
T is sufficient for this mixing. It is necessary to have a 
ready means of control for both the gas and the oxygen. 



[Chap. V 

If the latter is under high pressure, a reduction valve in 
conjunction with a needle valve gives the best regulation. 
A combination of needle valves and T which has been found 
to give satisfactory service is shown in Fig. 2. 

In using such a torch, care should be taken in lighting to 
turn the gas on first, light it, and then gradually turn the 
oxygen on until the proper flame is produced. To extinguish 
the flame, turn the oxygen off slowly and then the gas. Dis- 
regard of this procedure may 
result in a backfire into the 
fine but usually does little dam- 
age except to sensitive nerves. 
The described torch is a use- 
ful adjunct to any laboratory, 
especially when supplied with 
tips of various sizes. It is 
ideal for working Pyrex glass 
as well as quartz. When 
quartz fibers are being made, 
the torch is held by a clamp so 
that the flame is vertical. 
Indispensable in the working of small pieces of quartz is a 
small torch shown in Fig. 1, identical with the larger one 
except for size, and using the same gases, which are controlled 
by separate fine needle valves. The best metal tubing for 
this torch is brass or copper -^ inch in internal diameter. 
The gases are led from the mixer to the torch by ^-inch 
rubber tubing. Small volumes throughout are important, or 
much time will be wasted in waiting for a change of gas 
mixture to arrive at the tip. The tips should be inter- 
changeable and should have openings of from 0.05 to 0.2 mm 
in diameter. A slight modification of design (illustrated) 
permits the torch to be mounted and manipulated by me- 
chanical means. The usefulness of this small torch will 
become apparent later. 

In measuring the sizes of fibers, an ordinary microscope 
equipped with a scale in the eyepiece and having a magnifica- 

scale of inches 

Fig. 2. Combination of needle 
valves and mixer. 

Chap. V] 



50 to lac^u quartz prongs 


tion of from 300 to 1000 is very useful. With some experi- 
ence the sizes of fibers can be judged to within 20 to 50 per 
cent by the amount of scat- 
tered light, the way they 
weave in the air, and so forth, 
but in many cases the diam- 
eter is important, and an ac- 
curate means of determining 
their size is invaluable. 

After blowing out a fine fiber, 
two places are marked, and 
the position of the intervening 
portion is thus determined by 
small tabs. Dennison's No. 
251 tabs are recommended. 

In many instances one 
works with fibers from a few 

centimeters to 10 or even 20 cm in length. In these cases 
the fibers are mounted on a two-pronged fork. This is 
easily made as shown in the sketch, Fig. 3. The end of 

Fig. 3. A simple fork used for 
holding fibers while they are being 

fiber being worked 

hard wax 

J?P^ ^ T diffe 

: «50-i5Oyu * — -*\ angl 

"NJ quartz prongs! 

handles -should be 
attached at 

- /i6 w i res 

Tie" round 
rod 5# long 


A brass 
steel spring 


Fig. 4. Adjustable prong fork. 

each prong is drilled, and a piece of quartz (50/z to 150/x) is 
put in with hard wax. The reason for the quartz tips is to 



[Chap. V 

allow some freedom to the fiber, since the quartz tips will 
bend if the fiber is pulled one way or the other. Rigid 
supports result in many more broken fibers. The fiber is 
fastened to the tips with a small piece of hard wax. 

In cases in which one fiber is melted to another, each will 
shrink, and the quartz will gather at the junction. It is 
necessary then to have two forks, each with movable prongs. 

wooden handled %2 brass wire* 
to toy transformer *26 Chromel wireV 

Fig. 5. Hot-wire holder. 

The fork designed according to Fig. 4 has proved very satis- 
factory. If the handles are attached at different angles, the 
two forks can be worked together more easily. 

A hot wire mounted as in Fig. 5 has many uses and is 
especially valuable in melting small pieces of wax. The 
resistance wire can be any one of several, such as platinum, 
German silver, Chromel, Nichrome, and so forth. It should 
be 24-26 B and S gauge. A toy transform Br with variable 
voltage of from 1 to 6 volts is convenient for controlling the 
temperature. A foot switch is very useful, since both 
hands may be occupied when the heat is wanted. 

In testing for conductivity of quartz fibers which have a 
coating of metal, a probe (see Fig. 6) with a fine platinum 

""^Bakelite hand 
to galvanometer 

metal cap 
and source of low voltage *36 platinum wire 

Fig. 6. Platinum probe for testing conductivity of metal-covered fibers. 

wire tip finds a use. For such testing high voltages should 
not be used, since the resulting sparking will remove the 
metal from the fiber around the point of contact. Several 
volts applied through a 100,000-ohm resistance and a low- 

Chap. V] 



sensitivity galvanometer will be found satisfactory for 
qualitative work. 

Waxes are indispensable in fastening fibers either tempo- 
rarily or permanently. For general use Dennison's hard red 
wax, DeKhotinsky wax, or flake shellac is recommended. 
If the wax is holding in place two or more fibers which are to 
have a metal evaporated or sputtered onto them, one of the 
latter two waxes should be used and heated until polymeriza- 
tion takes place, resulting in a material either difficult or 
impossible to melt. Otherwise the heat developed during 
the process of depositing the metal may cause the wax to 
soften and the fibers to be displaced. 

y/&" bra^s rods r/ ^ N . 

lOO^i quartz fiber 
waxed into end with 
sealing wax 

needle held as above 
hairspring tweezers 

hairspring tweezers 
with clip to hold 
them closed 

man i cure scissors 
for cutting; fibers 

edge of one should be niched 
to prevent fibers from slipping 

Fig. 7. Various instruments useful in fiber work. 

In case it is necessary to hold a fiber temporarily and to 
maintain its desirable qualities, a wax must be used which, 
when heated, will completely disappear and not react in any 
way with the quartz. None of the products sold as waxes 
serve the purpose. An organic chemical which has the 
desired properties is diphenylcarbazide. It usually comes in 
powdered form and should be as pure as possible and espe- 
cially free from inorganic materials. 

In handling small pieces of wax, holding fibers, bending 
quartz fibers, and so forth, a piece of quartz 100/x in diameter 
and 2 to 3 cm long, waxed into the end of a metal rod, is very 
useful. (See Fig. 7.) It will also be found that a needle 



[Chap. V 


mounted in the end of a metal rod has many uses. It is 
recommended that several such quartz and needle holders 
be available. 

When working with small objects, tweezers of various 
sizes are very convenient. These can be obtained from 
jeweler's supply houses or from most houses supplying 
scientific apparatus. For very fine work, watch-hairspring 
tweezers such as #3C made by Dumont & Fils, Switzerland, 
are recommended. Also valuable in cutting fibers are small 
scissors. These may be a good grade of manicure scissors 
or dissecting scissors used in biological work. A nick should 

be made in one blade to prevent 
large fibers from slipping. If the 
scissors are guided by mechanical 
means, small fibers (up to 40/*) 
can be cut off as little as 0.01 mm 
at a time under a microscope. 

In most fiber work it is neces- 
sary to fix the position of the fiber 
with some accuracy. Small tri- 
pods with adjustable feet, to- 
gether with clamps and rods, as 
shown in Fig. 8, will serve to hold 
the various forks, needles, and so 
forth, used in the process of mount- 
ing the fibers. It is very difficult 
to hold a fiber still enough by hand, and it is always best 
to take advantage of mechanical devices wherever possible. 
Very small fibers (Ijj, and less) can be easily seen by 
scattered light against a black background. Black velvet 
is one of the best. If the diameter of a fiber is to be measured 
under the microscope, a light background is needed; the 
scattered light against black gives a false impression of the 
size, since the actual outlines of the object cannot be seen. 
To put a conducting coat of metal on quartz, any one of 
several methods can be used. The simplest, and one which 
is satisfactory for fibers down to 20/x in diameter, is to bake 

at I east one 
adjusting screw 

Fig. 8. Support for holding 
work or fixing the position of 

Chap. V] 



is rotated 
during evap 
oration by 
waving a 
near it 
outside the 
bell jar 

this coil for 
coating upper 
sides of 


coated with 
the metal 
to be evap- 

this coil 
for coating 
under sides 
of work 

Fig. 9. 

This arrangement allows the evaporated metal to be deposited on 
all sides of the work. 


the metal on, using any of the good china paints. Most of 
the noble metals — for example, platinum, gold, iridium, and 
so forth — can be obtained in this form. The paint is made by 
dissolving one of the metal salts in an organic liquid. China 
painters use this on their dishes and fire them to 700°C. 
The organic material disappears, and the metal compound 
decomposes, leaving behind a uniform coating of the metal. 
The thickness for each coat may vary from 0.05ju to 0.15^, 
depending on the thickness of the original paint. Very 
adherent, electrically conducting coatings can be applied to 
glazed porcelain, glass, quartz, and so forth. The hot wire, 
held under small pieces of quartz fibers covered with these 
solutions, will bake them in a few seconds. If an attempt is 
made to treat small fibers in this way, it will be found that 
the solution collects into small drops along the fiber, and a 
disconnected coating results when it is baked. 

Sputtering or evaporating the metal on are the most satis- 
factory methods and have the advantage that conducting 
coats can be applied to fibers of any size. In general it is 
desirable to arrange to coat the fibers on all sides. Evapora- 
tion is the easier and simpler of the two methods. (See Chap- 
ter III.) A suitable apparatus for this is shown in Fig. 9. 

In working quartz it is absolutely necessary to use dark 
glasses to protect the eyes. Besides the brilliant glow, which 
in itself is bad for the eyes, the light is very rich in ultra- 
violet, which is especially harmful and may cause blindness 
through long exposure. The glasses should be gray in color, 
preferably, and have a transmission of from 10 to 20 per cent. 
Ordinary glass will cut out the ultraviolet, so that inexpensive 
dark glasses will suffice. 

The writer has used for some time a set of three micro- 
manipulators. Each has a three-jointed arm, which allows 
complete freedom in determining the position of the fiber. 
For fine adjustment, micrometer screws with divided heads 
give accurate motion in three mutually perpendicular di- 
rections. The accompanying illustration, Fig. 10, shows 
one of the three manipulators. 



tubular socket 
for implements 

Fig. 10. Micromanipulator. 



[Chap. V 

Although much of the simpler fiber work can be done with 
the unaided eye or with a magnifying glass, for fine work in 
which accuracy is important and ease of working is desired 
a binocular microscope with a magnification of 15 to 20 can 
be strongly recommended. Such a microscope not only 
gives stereoscopic vision but when used properly results in 
little, if any, eyestrain. A scale in one eyepiece allows 

fiber storage >^ 
micromanipulators /<f^ 

on an 
adj ustable 

manipulator * 
SLsnrmll rubber^^ 
■tube may be 

to foot switch 
for hot-wire tool 

Fig. 11. Complete assembly for working quartz fibers. 

measurements to be made. Lighting from several direc- 
tions is desirable to provide proper illumination on the work 
in all positions. 

A complete setup of the major equipment used by the 
writer in quartz fiber work is illustrated in Fig. 11. The 
black glass base permits the fine fiber to be seen easily by 
scattered light. When the actual outlines of large fibers 

Chap. V] 



are to be seen, a piece of white paper is placed on the glass 
base and used as a background. 

Making fibers. A convenient size of stock quartz rod is 
3 to 4 mm in diameter. Smaller rod than this is apt to break 
when the larger fibers are being drawn and is not easily held 
in the hands. Larger rod becomes more difficult to melt. 

The first step in making a fine fiber is to draw one from 50^ 
to 100/jl in diameter. (See Fig. 12.) Two pieces of stock 

lorgfe hand 
torch supported 
on .stand 

ffber 50 to 100 /u in diameter 
i& to. 30 inches long 1 

Fig. 12. The first step in making a small fiber is to draw a larger one. 
very hot flame is used. 

quartz of convenient length are held in the hands. The 
oxygen-gas flame is adjusted to maximum heat; that is, 
both the oxygen and gas are increased, especially the oxygen, 
until a hissing flame results, and the small cone just over the 
opening in the torch tip has shortened until its height is 
perhaps two or three times its width. The hottest portion 
of the flame is just above this small cone. The ends of the 
quartz rod are melted together and then pulled apart a short 



[Chap. V 

distance, so that the connecting soft quartz is perhaps 1 mm 
in diameter. This portion, when held in the hottest part 
of the flame, will become quite soft. The quartz rods are 
then quickly removed from the flame, and at the same time 
the two pieces held in the hands are separated rapidly to a 
, distance of several feet. 

The hotter the narrow sec- 
tion of quartz and the faster 
the drawing, the smaller 
will be the resulting fiber. 
Fibers down to 20/jl can be 
drawn in this manner. 

To make a smaller fiber 
from the larger one, the 
procedure is as follows: 
Break the connecting fiber 
produced in the above 
drawing process so that a 
section of 8 to 10 inches is 
left on each piece of quartz 
stock. This section should 
be stiff enough to support 
itself in a vertical position. 
Now adjust the flame by 
turning the oxygen par- 
tially off, so that a steady 
flame about 15 to 20 inches 
long is produced. The cone 
above the tip will lengthen 
to several inches. Holding 
the quartz stock so that the attached fiber is vertical, move 
it into the vertical flame as illustrated in Fig. 13. The 
whole length of the fiber will glow uniformly. If the tem- 
perature of the flame and the size of the fiber are right, the 
fiber will gradually begin to lengthen, slowly at first and then 
more rapidly as it becomes smaller. Finally, the upper sec- 
tion of the original fiber will go quickly toward the ceiling. 

Fig. 13. The second step in making a 
small fiber is to blow out the larger fiber 
by holding it in a long, vertical, relatively 
cool flame. 


As soon as this happens, the lower end should be removed 
from the flame. A careful examination will reveal a fine 
fiber joining the two ends of the original, perhaps 3 to 6 feet 
long. Sections of it can be seen in scattered light. Place 
a small tab on one part of the fine fiber with one hand while 
holding the stock quartz (to which the other end of the fiber 
is attached) in the other. The position of the intervening 
portion is now determined, so that other tabs can be stuck 
on and suitable lengths removed. Each end of each length 
will thus have a small tab attached. These fibers are then 
stored in a clean container in which the air is kept dry. 
(See Fig. 14.) 

The size of the resulting small fiber will depend on a num- 
ber of factors. Chief among these are the size of the original 
fiber, the temperature and size of the flame, and the time 
intervening between the disappearance of the top of the 
original fiber and the removal of the lower end. Some 
practice is necessary to secure fibers of a desired size. It will 
be found that fibers produced in the above manner are 
straight and of quite uniform diameter for some distance 
on each side of the center. 

A few cautions are necessary if good fibers are to be had. 
The basis of all of these is cleanliness. Much of the dust on 
objects around a laboratory and floating in the air is inor- 
ganic. If a fiber is heated where a piece of dust has settled, 
the metallic salts form silicates and in general completely 
spoil the surface, and for that reason the fiber also, at the 
point of contact. It is a general rule that no part of a fiber 
which ultimately is to have any stress applied should ever 
touch anything except those materials which are softer than 
the quartz and will not react with it. This may seem to be 
a stringent requirement, but in reality the fiber can always be 
handled by its ends, which are eventually discarded. 

If the original large fiber shows any bright spots when put 
into the flame, it should be discarded. In general, this is the 
best test for dust that can be applied. Dust will immediately 
show itself by causing a bright spot, and the fiber can be 


discarded forthwith; if there is no dust on the fiber, it will 
not be harmed by heating. This test can be made with fi- 
bers from 10 ix to lOO^t with an ordinary Bunsen burner. 
For smaller ones the small torch using a pure gas flame should 
be used. In each case the fiber should be under some ten- 
sion to keep it straight. 

If the size of the fiber is to be measured with the micro- 
scope, it is usually sufficient to take a sample from each end 
and take the mean diameter. The sample is placed on a 
piece of glass, which in turn is placed on the microscope 
stage and viewed by transmitted light. To find the fiber in 
the microscope the following procedure is valuable in saving 
time: Have plenty of light passing through the optical 
system. Raise the objective until it is several times the 
working distance from the object. Remove the ocular. 
Move the glass on which the fiber is lying until, by looking 
down the microscope tube, the reduced image of the fiber is 
seen. Adjust the position of the fiber until its image ap- 
pears approximately in the middle of the objective. Now 
move the objective down until the image begins to spread. 
When it appears to cover the objective completely, the ob- 
ject is near the focus, and on replacing the ocular, the image 
should be in the field of view. 

After working with fibers for a while, one can judge their 
size by the amount of scattered light, the amount of weaving 
in the air, how much a fiber of a given length sags under its 
own weight, the radius of curvature when hung over a needle 
with a tab on one end, and so forth. These methods are 
good to from 20 to 50 per cent, except for fibers below 1/x 
to 2^. 

Another method for drawing fibers has been described 
by Boys. 7 It consists in pulling the two pieces of quartz 
apart very rapidly by means of a projected arrow. Long 
fibers down to 10/* of very uniform diameter can be produced 
in this fashion. The hotter the quartz and the faster the 
arrow is shot, the finer will be the fiber. 

7 Ibid., Volume III, page 696. 

Chap. V] 



The care and preservation of small fibers. When a fiber 
has its two ends marked with tabs, it should be hung in a 
clean, dry container. A crosspiece at the top of the con- 
tainer, on which are small pieces of soft wax or beeswax, 
serves as a hanger. The top tab is pressed into the wax, 
and the lower tab keeps the fiber from weaving around and 
touching things. 

The container should be 10 to 12 inches deep, airtight, 
and preferably made from glass. It should be clean and 
contain a good drying agent 
— either phosphorous pen- 
toxide or anhydrous potassi- 
um hydroxide. A convenient 
container is made from an in- 
verted bell jar with a plate- 
glass top as shown in Fig. 14. 
Fibers deteriorate in moist 
atmospheres, but can be pre- 
served for months with no 
change in breaking strength 
if kept clean and dry. 

Some useful techniques 
in fiber work. Straightening. 
Fibers from 10/* to 500/* can 
be quickly and easily straight- 
ened by hanging a weight on the lower end and running a 
Bunsen burner flame up and down the piece several times. 
The weight should be somewhat less than that necessary to 
elongate the fiber appreciably under the heat of the flame. 
A small Dennison tab is sufficient for fibers 10/* to 50/* and a 
|-inch tab for those between 50/* and 500/*. For fibers 
from 4/t to 10/* a small Dennison tab should be cut in two 
and the small torch burning pure gas used for heat. 

Bending. Fibers from 40/* on up are best bent by hang- 
ing a weight such as a tab at one end, holding the fiber 
at the proper angle, and applying the heat locally with a 
small torch burning oxygen and gas. The piece between the 

Preserving quartz fibers. 



[Chap. V 

flame and the tab will fall to a vertical position as shown 
in Fig. 15. 

Fibers between 1/* and 40/* are best bent over another 
piece of quartz. A weight such as part or all of a small tab 
or a small piece of wax bends the fiber over the larger piece 
of quartz (100/x or less). A pure gas flame applied with the 
small torch at the contact of the two fibers will bend the 


4(Yu fibev 
or larger 

Fig. 15. Bending large and small fibers. 

smaller one over the larger. The flame should not be ap- 
plied longer than is necessary, or the two pieces of quartz 
are apt to stick together. 

Drawing and shrinking. If one end of a fiber is attached 
to a screw-controlled sliding mechanism, such as the movable 
prong fork described earlier, a portion of it may be readily 
drawn down to any desired size by applying a flame with the 
small torch and gradually screwing out one prong. 

Soft quartz has a high surface tension, and fibers tend to 
shrink when heated. The heating is done with the small 
torch. It is necessary to have a properly adjusted flame. 
A compromise must be made between a hot flame with swiftly 
rushing gases, which readily melts and blows the fibers 

Chap. V] 



apart, and a cooler flame, which will not soften the quartz 
sufficiently. The ideal is reached when the tendency to blow 
away is overcome by the tendency to pull together due to 
surface tension. The fiber is heated in a slackened condition, 
and as the shrinkage proceeds it is fed by the movable prongs. 
A torch tip with a hole about 0.1 mm in diameter is perhaps 
the best. With some practice a fiber may be locally enlarged 
to many times its previous diameter. (See Fig. 16.) 

fiber dmws 
itself taut 
and thick- 
ens in the 

fiber is 



fiber is 



Fig. 16. Shrinking a small fiber. 

Joining one fiber to another. When the above technique 
has been learned, the joining of two fibers crossing one an- 
other becomes simple. Each shrinks to the common junc- 
tion, forming a joint which is stronger than any other portion. 
For this work it is necessary to use two of the forks with 
movable prongs, gradually feeding in the quartz as the joint 
grows in size. 

Joining a fiber to a larger piece of quartz. If the larger 
piece is too large to melt locally with the small torch, 
a "teat" is put on at the proper place with a larger torch 
and then drawn down to a fine point. The fiber, mounted 
on the fork, is placed next to this teat, and heat is applied 
to the teat. Upon softening, the larger piece of quartz draws 
the small fiber in by surface tension. Straightening of the 



[Chap. V 

fiber near the junction is done by heating with the small 
torch burning pure gas when the fiber is under a slight 

With care, fibers as small as lju in diameter can be melted 
to other fibers or larger pieces of quartz. 

Drawing an oval fiber. The tip of each piece of the stock 
quartz is heated in the oxygen-gas flame so that only the 

Fig. 17. Making a flat fiber. 

very end becomes soft. With the axes of the two pieces 
held parallel, the ends are brought together and immediately 
separated at right angles to the axes of the stock quartz, 
and at the same time they are removed from the flame. 
(See Fig. 17.) Only flat fibers larger than 30/* to 40/z can 
be produced in this fashion. They are useful in vibration 
types of pressure gauges in which the motion is to be limited 
to one plane. 


Drawing flat tubing. In some cases quartz is useful in 
making the Bourdon type of pressure gauge. If a long piece 
of flat tubing is made into a spiral and a mirror and scale are 
used to measure the change in angle, such a gauge becomes 
an accurate means of measuring moderate pressures. One 
way to produce long pieces of elliptically shaped thin-walled 
tubing is to use two large torches as cross-fires and to heat 
^- to }-inch quartz tubing without rotation. Heating should 
continue until the walls nearest the flame are quite soft. 
The tubing is removed from the flame and rapidly pulled to 
3 or 4 feet. If heating has not been sufficient, the elongated 
occluded bubbles will cause the resultant tubing to be brittle. 
It is, in fact, a good procedure to work the heated section by 
alternately enlarging and contracting it with internal 
pressure before drawing. The oval tubing is bent into the 
desired shape with a moderately hot flame. 

Making electrometer suspensions. Quartz fibers make ideal 
suspensions for electrometers. The most satisfactory way 

g^=. - -. ~ «_-^ 

Fig. 18. Design of quartz fiber support used in the Dolezalek and Compton 
electrometers. The whole is made from fused quartz, upon which is deposited 
a coating of metal, for example, gold or platinum. 

of making the suspensions consists in joining the ends of the 
fiber to two larger pieces of quartz by melting them together 
with a small torch. In many cases these larger pieces are 
bent into small hooks, and then the whole is made conducting 
by evaporating or sputtering gold or some other metal on it 
as represented in Fig. 18. In cases in which hooks cannot be 
used, the larger quartz is left straight and is cemented into 
place with a hard wax such as DeKhotinsky's. Contact is 
made by attaching a fine wire to the quartz with hard wax 
before the fiber is coated with the metal. The wire is later 
soldered to the metal pieces of the electrometer. 

The method of soldering the metal-coated fibers does not 
produce a suspension as permanent as with the methods 
described above. The gold is apt to amalgamate with the 



[Chap. V 

solder and result in a poor contact between the main portion 
of the fiber and the solder. 

Another method of fastening fibers to metal parts and at 
the same time making an electrical contact is to use colloidal 
graphite. A small drop is placed at the proper point, and 
in a short while the water will evaporate, leaving a strong 
conducting joint. 

Quartz is very convenient for making various types of 
electroscopes. It is not only good for the moving parts but 
is used uncoated for insulation. 8 

Mounting cross hairs in optical instruments. Fibers made 
from quartz surpass any other material for cross hairs. 
Owing to the refraction of the light by the fiber, it appears 
black as seen in a bright field. Its essential smoothness, 
freedom from dust, uniformity of size, straightness, and the 
fact that it can be drawn to any desired diameter make it 
especially valuable. 

The mounting is first prepared by melting hard wax onto 
it at the desired points. The fiber is mounted on a fork and 

diaphragm rind 

.ends clipped 

cross hair 

Fig. 19. Steps in mounting cross hairs for microscope and telescope eyepieces. 

lowered into position. A hot wire brought near the wax 
where the stretched fiber rests will allow the fiber to sink in 
and become firmly attached. The various steps are illus- 
trated in Fig. 19. 

Torsion balance. For objects weighing less than 1 mg 
the torsion balance becomes very useful. It is not difficult 
8 See Chapter VI, "Electrometers and Electroscopes." 

Chap. V] 



to make a balance having a sensitivity of 10 -7 to 10 -9 
g/div. without the use of mirrors or microscopes. A simple 
calculation will show the size of fiber necessary for the 
specific requirements. The crossarm should be statically 
balanced. The amount of twist of the fiber is conveniently 
read from a divided head. 

The balance may be calibrated by weighing on an analyti- 
cal balance a long section of fine wire such as 40 B and S 

wooden cose with gleuss ^ 

windows to cut down convection 

dikl for 

fiber and 


object to I 
be weighed 

Fig. 20. Simple design of a quartz microbalance. 

gauge copper, 2-mil nickel, or smaller if needed, and cutting 
from this piece samples of a given length. Usually ten 
samples will give a probable error of less than 1 per cent in 
th?, calibration. If the tension in the torsion fiber is kept 
constant with a quartz bow, it can be assumed with much 
accuracy that the twist is proportional to the weight. Since 
<\>r/l is the surface strain, where r is the radius of the torsion 
fiber, <f> the angle of twist in radians, and I the length twisted, 


and since the maximum value of this is about 0.05, the maxi- 
mum load which the balance can handle is easily computed. 
A simple design of such a torsion balance is shown in Fig. 20. 

If all the joints are made of fused quartz, there need be 
no fear of a changing "zero," since the limit of elasticity 
coincides with the breaking point. 

Other uses of quartz. Quartz rod or fiber is often used as 
a carrier of light — visible, ultraviolet, or infrared. In- 
ternal reflections keep the light inside the quartz and permit 
it to be led around corners, provided the corners are not too 

In many cases in which accuracy in maintaining shape or 
position is important, quartz finds a use. All metals change 
their dimensions with time, especially when under strain. 
This change can be lessened by thorough annealing, which 
consists in subjecting the metal alternately to temperatures 
above and below room temperature. In extreme cases this 
treatment may take days or weeks. Annealed fused quartz 
does not suffer from changes in dimensions, since the flow 
under strain is less than 10 -3 of that for metals. 

Fused quartz is finding increasing uses in lamps of various 
kinds in which the transmission of ultraviolet light is im- 
portant. For the same reason many photoelectric cells 
are made from quartz. 

Although the above does not pretend to be an exhaustive 
list of the uses to which fused silica can be put, it is hoped 
that the reader will gain some idea of the usefulness of this 


Electrometers and Electroscopes 


H. V. Neher 

Definitions. It is not always clear just what distinguishes 
an electrometer from an electroscope, and there seems to be 
some confusion in the literature. For purposes of this dis- 
cussion the following distinction will be made: An electro- 
scope is an electrostatic measuring device in which only one 
potential difference is needed for its operation. Electrom- 
eters, on the other hand, need auxiliary potentials for their 
operation. The familiar gold-leaf electroscope and the 
quadrant electrometer are respective examples. 

General theory. 1 Expressed in terms of Maxwell's 
coefficients, the electrostatic energy of any system of con- 
ductors at potentials Vi, V2, • • • V n is given by 2 

W = i(diTV + 2c 12 7 1 7 2 +...) = i?QV, (1) 

where the coefficients of capacity, Cn, c u , Cu, and so forth, 
are given by 

Ql = CllVl + C12V2 + . . . Cl n V n , ' 

Qn = CniVi + c n2 V 2 + . . . c nn V„ 


the Q's being the charges on the conductors. The coefficients 
have the properties that c tJ = c n . c n is the charge which 
is on conductor 1 when all the other bodies are grounded and 

1 This treatment follows, in general, that given by Hoffmann, G., in Hand- 
buch der Exp. Physik, X, 42 (1928). Wein, W., and Harms, F., editors; 

2 Jeans, J. H., Mathematical Theory of Electricity and Magnetism, Fifth 
Edition, page 95. New York: The Macmillan Company 



1 has a potential of unity. Similarly c i2 is the charge in- 
duced on 1 when all the other bodies are grounded and unit 
potential is placed on 2. 

If the c's are functions of a coordinate £ and the V'& are 
kept constant, then the force or torque tending to in- 
crease £ is 

f = ^ + *f™ + ---> (3) 

In all instruments this electrical force or torque is balanced 
by some restoring force. If we know how the c's and the 
restoring force depend on £, then with the help of Eq. 3 we 
can solve for the voltage sensitivity. 

Since most electroscopes and electrometers are used to 
measure electric charge, it is the charge sensitivity in which 
we are interested, although the voltage sensitivity is the 
more easily measured. 

Referring bad: to Eq. 2, let i be the moving system. The 
charge sensitivity, S Q , is given by 

1 dQ t dVi T/ den . 

^ = ^ = cir w + 71 ^" + - • ■' (4) 

since both the c's and the V's are, in general, each functions 
of the coordinate. 

Applications to electroscopes. The above theory when 
applied to electroscopes becomes very simple. In this 
instance we have but two conductors, usually one completely 
surrounding the other. Let the case be grounded. Then 
Eq. 1 becomes merely W = §cF 2 , where c is the capacity 
of the electroscope system to the case, c must be a function 
of the displacement and in most instances can be considered 
a linear function, that is, c = c Q -\- &£. For this case Eq. 3 
becomes dW/dt; = |&7 2 . If this is balanced by a force 
proportional to the displacement, then &£ = J67 2 , and the 
voltage sensitivity is 

Sv ~w-j^' (5) 

From Eq. 4, the charge sensitivity, S Q , is given by 


Now dV/dt- is given by Eq. 5, and dc/d£ = b if c is a linear 
function of £. Therefore 

*° " c& + F 2 b 2 Kl) 

For most electroscopes, 6 2 F 2 is small compared with ck, so 
that we can usually write 

Sq -^-t' (8) 

This last relationship between the voltage and charge 
sensitivities is the usual assumption made when working 
with electroscopes. 

However, b 2 V 2 need not be small compared with ck, and it 
is interesting to see what follows in such a case. It will be 
seen that Eq. 7 has a maximum value when Vo 2 b 2 = ck, and 
under these conditions 

osy max . = kck)-«* = l 

2 V ' 2V b 

The effective capacity has increased to 


Cett - = df = 2c ' 

Any further increase in the voltage sensitivity results in a 
more rapidly increasing capacity and a decrease in charge 

These conditions more aptly apply to electrometers and 
will be discussed in that connection in the following section. 

Applications to electrometers. All electrometers can be 
considered as made from three conductors, two of which are 
stationary and usually similar, while the third is movable. 
All three are connected electrically to the outside of the in- 
strument. We shall assume in the following discussion that 


the charge or potential to be measured is applied to the 
movable system, while the two stationary parts are main- 
tained at equal and opposite potentials. This arrange- 
ment is not necessary, and in general the following discussion 
holds equally well for the case when the charge or potential 
to be measured is applied to a stationary part and the mov- 
ing system is kept at the high potential. It is further 
assumed that the electrical and mechanical zeros of the in- 
strument coincide. This last condition is fulfilled if, while 
the movable system is grounded with equal and opposite 
potentials applied to the stationary parts, no deflection takes 

Of the twenty-seven terms in Eq. 3, twenty-five are small 
or zero, as compared with the remaining two under the above 
conditions. If the moving part is symmetrical with respect 
to the stationary parts, then these two terms are equal, 
and Eq. 3 becomes m 

~W = VlVr df 

where c u is the capacity between the moving system and one 
of the stationary parts, which is at a potential Vi, while Vz is 
the potential of the moving system. 

Case I. In general, c i3 will be a complicated function of £. 
This is especially true in the case of most string elec- 
trometers, but with the Hoffmann and Dolezalek the de- 
pendence of the capacity on the displacement is approxi- 
mately linear. For these instruments c u = Ci 3 ° + bd, and 
the torque becomes t — v Vh /q\ 

In equilibrium, L = kd, where h is the torsion constant of 
the suspension. It will be noticed that Eq. 9 is symmetrical 
in Vi and 7 3 , so that the sensitivity is the same whether the 
stationary or the moving parts are at the fixed potential. 
Assuming that the stationary parts are kept at a fixed po- 
tential, the voltage sensitivity is 

S r = f • (10) 


The charge on the moving system, 3, becomes 

Qs = cisVi + c 23 V 2 + C33F3. 

Now the moving system is connected to a suspension which 
in turn is connected to an external capacity. Let the sum 
of these two capacities be denoted by c e . If a charge q is 
placed on the system, then Qs = q — Vzc e , and we have 



S a 


dd Cu 69 

q = C13V1 + C23V2 + (c e + c 33 )F 3 

T r dCi3 . dV 2 , T/ dC 2 3 , / , N 5F 3 . Tr dc 3 3 

Now c 33 , to the desired approximation, is not dependent on 
0, and dcu/dd = - dc 2 z/dd = b. Also V 2 = - Vi and 
d7i/d0 = a7 2 /a0 = 0. The 
charge sensitivity becomes 


S a = 

2V3* + (c. + c 33 )fc 

(c e + c 33 ) may be lumped into 
one quantity c, which is the 
electrostatic capacity of the 
electrometer and external sys- 
tem when the stationary 
pieces are grounded. Hence 

Sa = 2VW + ck' (11) 



This has a maximum value 
at 7 = (l/6)(c/c/2) 1/2 , and 
this is the value of Vi which 
should, if possible, be used on 
the binants or the quadrants, 
as the case may be, if high 
charge sensitivity is wanted. This makes (S a ) mta . = J (2ck)~ 1/2 = 
1/(4 Vob). Any further increase in Vi will increase the volt- 
age sensitivity but will decrease the charge sensitivity. The 
behavior of the total capacity, voltage sensitivity, charge 

Fig. 1. The total capacity, voltage 
sensitivity, charge sensitivity, and 
period (ordinates) of an ideal elec- 
trometer in terms of their values at 
the optimum value of the plate 
potential. The plate potential is 
measured along the abscissa, and its 
optimum value is taken as unity. 
See W. W. Hansen, R.S.I., 7, 182 


sensitivity, and period, as the auxiliary potentials are 
changed, is shown in Fig. 1. 

Experimentally, the proper value of Vi is easily deter- 
mined, as will be seen from what follows. The effective 
capacity is -q , 


C e fl. — 

= 2c 


at the optimum value of V\. The procedure is as follows: 
Determine the electrostatic capacity of the suspension and 

moving system, together with any 
permanently connected capacity 
such as an ion chamber. Call this 
total c. Compare the deflection of 
the electrometer when a potential V 
is applied with that obtained when 
the same potential is applied 
through a known capacity c s as 
shown in Fig. 2. Then when the 
ratio of the deflections in the two 




cases is 

Fig. 2. A method for de 

termining the effective capac- q 2c A- C 

ity of an electrometer in -~ = -> 

terms of a standard capac- ^ c s 

Vi has the proper value. 
The period may also be computed. When there is no net 
charge on the electrometer and when potentials are on the 
binants or quadrants, let the system be deflected through 
an angle 6 by some external means and then allowed to 
vibrate. By Eq. 11, 

dV 3 


27,6 + c^ = 0, 

V, = - 


The electrical torque from Eq. 9 is therefore 


Now the resultant torque is 


the equation of motion becomes 

- k$, 


and the period is 

m = _ ( 2VW- 
dt 2 \ 



T = 2* J 2PW ^ jl 

where / is the moment of inertia of the moving system. 
Now at the optimum value of Vi, 2Vi 2 b 2 — ck, and hence 
the corresponding period is 

T = 2 W1 

2k (12) 

In other words the period has become 40 per cent less than 
when no potentials are applied. The maximum charge 
sensitivity in terms of the period, capacity, and moment of 
inertia of the system becomes 

(^)max. = |V)- 1/2 ' (13) 

Case II. If the coefficients c lz and c 2 3 are quadratic 
functions of the displacement, that is, if 

C13 = c u ° + bd + gp, 

the values of the voltage and charge sensitivities are as 
follows : 

bv ~ k - wm' 


by = 



Q V 1 (b + 2gd) 

2Fi 2 (6 + 2gd) 2 -4- ck 

The optimum value of Vi is given by 

= _j__MY /2 
Vo b + 2gd\2j 

This last is just what was obtained when the dependence of 
the capacity between the moving and stationary parts was 
a linear function of the displacement. 

The effective capacity at the maximum charge sensitivity 
is the same as in the simpler case, namely, twice the pure 
electrostatic capacity. 

The total torque may be written as 

£=_(*■+ 2^ _«!££. 

If g is positive, then the electrometer has what is known as 
positive control, while if g is negative, it has negative con- 
trol. In the latter case the net torque may become zero at 
some point of the deflection, in which case the instrument 
becomes unstable. This frequently occurs when the sensi- 
tivity is high and is especially true with string electrometers, 
limiting the useful range to deflections near the midpoint. 
The period becomes longer if g is negative and will become 
longer the greater the amplitude of vibration. 

It will be seen from the characteristics of the instrument 
in Case II that they are not so desirable as those in Case I, 
since they depend on the amount of displacement. How- 
ever, there may be other advantages which make instru- 
ments of the second type more desirable, such as portability, 
ease of operation, and so forth. It should be borne in mind 
also that the above theory contains many simplifying 
assumptions, and the actual behavior of the instrument in 
some cases may be quite different. The chief differences are 
due to (1) a more complicated dependence of the capacity, 
between the stationary and moving parts, on the displace- 
ment and (2) air damping of the moving system. It is 
important to realize that where electrical charge is to be 


measured, there is an optimum value of the potential 
applied to the stationary parts for which the charge sensi- 
tivity has either an optimum or a maximum value, and that 
there is experimentally an easy way to test for such a condi- 

Some types of electroscopes. The familiar gold-leaf 
electroscope is made either with a vertical stationary metal 
piece and a single strip of gold leaf fastened near the top, 
or with two gold leaves mu- 
tually repelling each other as 
shown in Fig. 3. The lead- 
in is insulated from the metal 
box with an amber or sulphur 
bushing. The capacity will 
be from 3 to 5 cm and the 
potential necessary to give a 
45° deflection will be from 300 
to 500 volts. When the leaf is 
observed with a microscope 
or a telescope, it becomes a 
quantitative instrument and 
will serve many purposes 
where high charge sensitivity 
is not important. The tech- 
nique of mounting the gold leaves will be discussed at the 
end of this chapter. 

The Wilson tilted electroscope, designed by C. T. R. 
Wilson 3 and G. W. C. Kaye, 4 is a hybrid of the electroscope 
and the electrometer. The narrow gold leaf in Fig. 4(a) 
hangs normally in a downward position, and is observed by 
means of a microscope with a micrometer ocular. A po- 
tential of about 200 volts is applied to the plate. This 
plate is adjustable; that is, it can be moved in or out along 
the axis of its support. The proximity of the plate and the 

3 Wilson, C. T. R, Cambridge Phil. Soc, Proc, 12, 135 (1903). 

4 Kaye, G. W. C, Phys. Soc, Proc, 23, 209 (1911). 

This instrument is made by Cambridge Scientific Instrument Company, 
Ltd., Cambridge, England. 

gold leaf 

Fig. 3. 

Two types of gold-leaf 


potential applied to it give an electrostatic control which 
tends to neutralize the effect of gravity on the leaf. 

Three cases in general may be cited for the voltage sensi- 
tivity as shown in Fig. 4(b): Case I, where there is little 
electrostatic control and the voltage sensitivity is linear 
over the entire scale; Case II, where the leaf is stable over 
the whole range but the electrostatic control is almost 
sufficient to neutralize the effect of gravity over part of the 



potential of leaf 

Fig. 4. Schematic diagram of the Wilson tilted electrometer and some 
typical sensitivity curves. 

range; Case III, where there is an unstable region and conse- 
quently two " zeros." Case II is the most useful, and if 
deflections are taken over the same regions of the scale, 
there is no trouble about nonlinearity. 

The Wilson tilted electroscope, while it may find a use in 
some types of work, has been largely displaced by more 
modern instruments, such as quadrant and string electrom- 

The Wiilf bifilar electroscope 5 has frequently been used in 
cosmic-ray work. It is well suited for a portable instru- 

5 This instrument can be obtained from E. Leybold's Nachfolger A. G. Koln- 
Bayental, Bonner Strasse 500, Germany. 


ment but must be read in a fixed position. As is the case 
with most electroscopes, not only the reading but also the 
calibration is affected by tilting the instrument. It is 
usually enclosed in an airtight ionization chamber, in which 
the gas pressure is often increased to increase the number of 
ions formed by a given radiation. The charge is renewed 
on the electroscope either by 

metal socket 
cemented to 
quartz rod 

fibers soldered 
with Wood's /\ 
metal -f 

quartz fibers 
5 to 10 cm 
long; and 
10 to 20JU 
in diameter 

clean quartz 
^ rod 

metal rod 
O.s mm in 
at end 

a mechanical arm working 
through an airtight bushing 
in the wall or, what is better, 
by an internal arm operated 
by an electromagnet. 

The construction of the 
electroscope proper is shown 
in Fig. 5. A clean quartz rod 
is cemented into the metal 
piece which holds a short 0.5- 
mm rod by means of a set- 
screw. The small rod is flat- 
tened'at the lower end. Two 
metal-coated straight quartz 
fibers from 5 to 10 cm long 
and from 10 fx to 20 jjl in di- 
ameter are cemented or sol- 
dered (with Wood's metal) 
to the flattened piece. The 
lower ends .of the two fibers 
are cemented side by side 
to an insulating quartz bow made from 10/z to 20/* fiber. It 
is essential that the fibers, when uncharged, hang parallel to 
each other. Means of straightening quartz fiber will be found 
in Chapter V. If shellac is used as a cement, there will be 
sufficient conductivity from the metal to the fibers. The 
potential of 200 to 400 volts is applied to the upper metal 
support. If the fibers are 20/z in diameter and 8 cm long, a 
spread of 3 mm will be produced by about 300 volts. The 
capacity will be in the neighborhood of 1 cm. 

scale m cm 

O 1 

l^-^ — I — 

clean quartz 

bow 10 to 20/; 


? 1 

Fig. 5. Wiilf bifilar electroscope. 


The plane of motion of the fibers should be perpendicular 
to the optical axis of the microscope, the necessary adjust- 
ment being made by rotating the piece held by the setscrew. 

It is possible to increase the sensitivity of the Wiilf electro- 
scope either by increasing the magnification of the micro- 
scope or by decreasing the diameter of the fibers. However, 
this is limited by the fact that the collecting potential for 
the ions should not drop too low, depending on the nature of 
the gas and its pressure. It is customary to have a collecting 
potential of not less than 100 volts. 

From Eq. 8 the charge sensitivity is 

Q _ S v 

W C 

and if n is the average number of ions per second collected 

from each cubic centimeter of 
the gas, then 

metal-coated quartz 
supporting bow > 

clean quartz 



metal socket 
cemented to 
quartz rod 



c „ d% 

n = — S v -r:> 
ev at 

where v is the volume of the ion 
chamber and e is the charge on 
the ions. The capacity c will 
vary with the spread of the 
fibers. The determination of 
c for different displacements 
amounts to the determination 
of b in c — Co + &£• 

Regener's electroscope is a 
single-fiber type shown in Fig. 6. 
The conducting quartz fiber 
or Wollaston wire is mounted 
near a metal piece and is held taut by a fine bow. The 
whole is supported by a quartz insulator, and the charge is 
renewed in a way similar to that used with the Wiilf type. 

Lauritsen has used a small quartz-fiber electroscope with 
much success not only in small pocket dose-meters for 

fiber or 

quartz bow 
same as 

Fig. 6. Regener's electroscope. 


X-ray work but also in measuring radiations found in nuclear 
investigations. Its outstanding feature is its simplicity. 
A wire is flattened at one end and bent over at right angles. 
A 5fx (0.005-mm) metal-coated quartz fiber about 6 mm long 
is cemented to this flat piece with shellac or colloidal graphite, 
making an angle with the wire support as shown in Fig. 7. 
A short piece of the same size fiber is cemented to the end of 
the longer fiber, at right angles to the plane of the wire and 
to the first fiber. This added piece is to form an index for 
viewing with a microscope. The wire support is mounted 
in an amber insulator, which in turn is mounted on the end 

amber support* metal wire flattened 
a \ at end > 

f la^dtoL I k d 3— £^ JL — 

quartz. T 

Ramsden microscope objective X coated 

eyepiece transparent forms an image of metal-coated 

J r scale the'T on the scale quartz fiber 

Fig. 7. Lauritsen's electroscope. 

of the microscope. When the electroscope is used inside 
an ionization chamber, contact is made by a movable arm to 
the base of the metal support. 

In order to obtain reliable readings on cosmic rays in air- 
planes, a torsion type of electroscope was developed in 1932. 
It was necessary to have a self-recording instrument of high 
sensitivity, the readings of which would not be affected by 
tilt or vibration of the plane. As far as tilt is concerned, 
this effect on the readings can be reduced to less than 0.001 
of the total deflection for a tilt of 90°. As for vibration, satis- 
factory readings have been obtained with the electroscope 
mounted within 3 feet of the engine in a pursuit airplane. 

A drawing of the electroscope is reproduced in Fig. 8. 
It is made entirely of fused quartz. The torsion fiber is 
stretched until its length is increased about 1 per cent. 
The crossarm is bent at right angles at one end and, in case 
high magnification is used, it is drawn down to a convenient 
size. A short bit of fiber serves as a fiducial mark. The 


shapes of the stationary parts combine to give a linear scale 
over most of the range of discharge. A piece of platinum 
cemented to the quartz with a polymerizing cement is the 
point at which a new charge is placed on the system. With 
a very small oxygen-gas flame all joints are fused together 
so that the whole system becomes essentially one piece of 

clean quartz 

gold coated J* 

below this point*' 

platinum collar 
to receive charge 

torsion fiber 

vane in 



fiducial points 

Fig. 8. Torsion type of electroscope. 

quartz. The system from the platinum down is covered 
with a conducting layer of gold. The vane is balanced by 
cutting off one end. For many applications this balancing 
need not be done with great care and, in fact, becomes rather 
delicate if a very fine torsion fiber is used. If too much is 
cut off, mass can be added by applying some thin gold china 
paint 6 and heating it with a hot wire. 
6 See Chapter V. 


In general it will be necessary to put a permanent twist in 
the torsion fiber. This can be done by forcing the vane 
beyond the stop through the desired angle, relieving the 
tension by pushing on the bow at the bottom, and heating 
the fiber at each end with a smal) pure gas flame. This will 
soften the quartz just enough. The twist, of course, must be 
put in before the system is covered with its conducting coat- 
ing of metal. The following illustration gives some idea of 
how much twist is needed: With a torsion fiber 5/jl in di- 
ameter and 12 mm long and a crossarm or vane 18 mm long, 
if a 30° twist is put in the fiber, the deflection will begin at 
about 200 volts and the sensitivity will be about 2 X 10 -3 
radian/volt. The electrostatic capacity will be about 0.5 cm 
and the charge sensitivity about 1.2 radians/statcoulomb. 

Assuming a rigidity modulus of 5 X 10 11 dynes cm -2 , the 
torsion constant comes out 1 X 10~ 2 dyne cm radian -1 under 
the above specifications, and b (see Eq. 5) has a value of 

b = £?* = 0.6 X 10- 2 cm radian- 1 

with V = 1 statvolt. Since b is a geometrical quantity, it 
will not depend on the size of the torsion fiber. The above 
relation may be used to get an approximation to the sensi- 
tivity for other values of the torsion constant k. 

If a very fine torsion fiber is used, in order to keep the 
collecting voltage up it may be necessary to twist the torsion 
fiber around one or more times. If the crossarm is not larger 
than 20^ m diameter, this can be done manually after the 
conducting coat has been put on by using a needle and forc- 
ing the ends through, between the torsion fiber and the 
main quartz support. 

Some types of electrometers. The Dolezalek quadrant 
electrometer 7 is perhaps the most common type and the 
most useful. The general plan of the instrument is shown in 

7 This instrument is made by many firms, including the Cambridge Scientific 
Instrument Company, Ltd., Cambridge, England, the Cambridge and Paul 
Instrument Company, Ltd., and E. Leybold's Nachfolger A. G. Koln-Bayental, 
Bonner Strasse 500, Germany. 


Fig. 9. It consists of a cylindrical box, or "pillbox," di- 
vided into four equal and insulated quadrants. Opposite 
quadrants are connected together. There are two ways of 
using quadrant electrometers. One is to keep the needle at 
the high potential with respect to ground and apply the 

quartz i 

hook made 
of flattened 
end of wire 



*28 wire 


opened to 
show vane 

i brass 

am bcF 

plan of quadrants 
o sccs.Se of cm 6 

«= I 1 I I . | ! | 

Fig. 9. Dolezalek quadrant electrometer. 

charge to be measured to one pair of quadrants while the 
other pair is grounded. The other way is to maintain one 
pair of quadrants at the potential + V and the other at — V 
and place the charge to be measured on the needle. The 
first method is illustrated in Fig. 10(a). V will usually be 
from 50 to 150 volts, depending on the desired sensitivity. 
For the second case the battery connections can be those 
shown in Fig. 10(b), where the main batteries furnish only a 



potential and slight adjust- 
ments are made by a potenti- 
ometer as shown. Or the main 
batteries may be placed across 
the high resistances R and R, 
and adjustments are made with 
the potentiometer R', as rep- 
resented in Fig. 10(c). The for- 
mer circuit has the advantage 
that the life of the high-volt- 
age batteries is essentially their 
shelf life, while the chief ad- 
vantage of the second is that 
| + Vi\ always equals | — FJ 
and that the mechanical and 
electrical zeros remain together 
once they are made to coin- 
cide. However, modern "B" 
batteries maintain a remark- 
ably constant potential at no 
current over long periods of 
time and have a very low 
temperature coefficient, so 
that in many cases the first 
circuit can be used. 

An approximation can be 
made to the value b for the 
Dolezalek electrometer in 
terms of the geometry of the 
instrument. It will be seen 
that the vane is of such shape 
that the capacity between the 
vane and the box, as a deflec- 
tion takes place, varies linearly 
with the change of angle. Let 6 
be the deflection, R the radius of the vane, h the distance of 
the vane from one side of the box, and d the depth of the box. 

■V t 


mm— urn 



(c) -W 

Fig. 10. Methods of applying the 
fixed potentials to the quadrant 


Then b is the increase of capacity between the vane and the 
conductor into which it moves per unit of angle, or 

torh(d - h) ' 

Since there is an equal and opposite vane on the other side, 
the total electrical torque, by Eq. 9, is 

±wh(d - h)' 

which at equilibrium is equal to kd. The voltage sensitivity 
is then 

Sv = 4*hkld - h)' (U) 

and the charge sensitivity is 

IwViRthdid - h) 
Q " (VJPdf + S**ckh*(d - A) 2 ' { } 

At the optimum value of Vi, 


4irh(d - h) /ckV 
RH V2 7 

and the maximum charge sensitivity becomes 

(^) max . = i(2ck)-^. 
The effective capacity at this sensitivity is, of course, 2c, 
where c is the total electrostatic capacity on which the 
charge q is placed. 

Eqs. 14 and 15 predict a constant voltage and charge 
sensitivity for given values of the potential on the quadrants 
or on the needle and for given geometrical conditions. 
Actually, however, this is not the case, for it will be found 
that as the potential on the quadrants or on the needle, as 
the case may be, is increased, the period gradually lengthens, 
and a value is finally reached at which the vane becomes un- 
stable at a certain point of the scale. This behavior is due 
to the occurrence of nonlinear terms in the expression for 
the capacity between the vane and the quadrants. By 
careful adjustment the importance of these terms can be 
diminished but never eliminated. 


In setting up the Dolezalek electrometer, the vane should 
not be too close to either the top or the bottom of the box, 
since small variations due to changes in temperature and 
so forth will change the characteristics. Also, irregularities 
in the vane may make important nonlinear terms in the 
capacity between the vane and the quadrants. Although 
the maximum charge sensitivity is not affected by this 
distance, the optimum voltage and the voltage sensitivity are 

The instrument is leveled until the piece holding the vane 
is in the center cf the circular hole in the top of the box. 
A grounding switch, which may be manually or magnetically 
operated, must be provided. Care should be exercised not 
to introduce variable thermal e.m.f.'s. 

The torsion head should be adjusted so that each half of 
the vane lies as nearly as possible symmetrically between 
two quadrants. With the vane grounded, small values of 
-f- V and — V are applied to the quadrants. Adjustment is 
made to the torsion head in the appropriate direction, so 
that whether a potential is on the quadrants or not, no mo- 
tion of the vane takes place. After the full values of d= V 
are placed on the quadrants, slight adjustments can be made 
with the potentiometer as shown in Figs. 10(b) and 10(c). 
This procedure is generally known as bringing the mechanical 
and electrical zeros together, and must be done with all 
forms of electrometers. 

In operation, the actual useful working value of the charge 
sensitivity will be about 

£ a = 1.3 X 10 4 div./statcoulomb 
= 0.4 X 10 14 div./coulomb 
= 0.6 X 10~ 5 div./electron. 

The corresponding voltage sensitivity will probably lie in 
the range ^ = 1QQQ tQ 150Q div#/volt> 

while the optimum value of the voltage applied to the 
quadrants will probably be between 50 and 100 volts on 


each side of ground, or if the high potential is placed on the 
needle, it will usually be between 100 and 200 volts, depend- 
ing, among other things, on the size of the torsion fiber. 

for raising' 
or lowering 

case with 



end view of 
vane showino() 
angle - & - 



hook made 
of flattened 
end of wire 

35 wire 





scale of cm 
Fig. 11. Compton adjustable quadrant electrometer. 

It is assumed in the above that a scale with 1-mm divisions 
is used at the customary distance of 1 m. 

The Compton electrometer 8 was introduced in 1919 by the 
two Compton brothers. 9 It is of the quadrant type but is so 

8 This instrument is made by the Rubicon Company, 29 North Sixth Street, 
Philadelphia, and by the Cambridge Scientific Instrument Company, Ltd., 
Cambridge, England. 

9 Compton, A. H. and K. T., Phys. Rev., U, 85 (1919). 


arranged that one quadrant can be raised or lowered with 
respect to the other three. Further dissymmetry is intro- 
duced by giving the vane an initial tilt. By proper adjust- 
ment of this movable quadrant the time consumed by the 
needle in returning to its initial position after a deflection 
may be lengthened (negative control) or shortened (positive 
control). The design of the instrument is shown in Fig. 11. 
The dissymmetry introduces additional nonlinear terms 




■ V 


f *>+ 











+ , 

+ /-(- 

374 45./ 



/ + / + 
17/ / 

J 1 Z2 J 


ao 40 





ZO 40 





60 %0 iOO 

Fig. 12. Typical curves of the Compton electrometer showing the effect 
of various degrees of positive (a) and negative (b) control on the voltage 
sensitivity. The curves in (b) were taken with a much stiff er fiber than those 
in (a). 

into the change of capacity as the needle moves, and an elec- 
trostatic torque is introduced which either opposes the torque 
of the suspension (negative control) or aids it (positive 
control). In the extreme case the action of the suspension 
can be more than completely neutralized, so that an unstable 
instrument results. This means that the voltage sensitivity 
can be made extremely high. Fig. 12(a) illustrates the 
relationship between the voltage sensitivity and the voltage 
on the needle for different degrees of positive control, while 
Fig. 12(b) shows the same relationship for various degrees of 
negative control. In this latter case a stiffer suspension was 
used. The circles on curves 6, 7, 8, 9 represent the highest 


sensitivity at which the zero of the instrument is sufficiently 
stable to allow satisfactory measurements to be made. 
The small figures above the curves of Figs. 12(a) and 12(b) 
represent the time required in each case for the needle to 
return to within 1 mm of the rest position after a deflection 
of 50 mm. Because of the small restoring torque and high 
air damping, the motion of the suspended system is aperiodic. 

Where extremely small potentials are to be measured and 
where the demand on charge sensitivity is not too great, the 
Compton electrometer is very suitable. However, the same 
voltage sensitivity could be achieved with the usual quadrant 
electrometer by putting in a suspension fine enough to give 
the same time of return to zero, provided that the moving 
system were equally as light as that in the Compton. In 
fact, it will be noted from the curves that with neither posi- 
tive nor negative control, shown by the straight line of 
Fig. 12(a), but with a fine fiber, voltage sensitivities can be 
obtained equal to those with a stiff fiber and large negative 
control. This high voltage sensitivity is not always useful 
when measuring electric charges, which is the main purpose 
of electrometers, for not only does the instrument become 
very sluggish, but drifts become bad. Also, for a given time 
of return from a given deflection the charge sensitivity has a 
maximum value. 10 Wolf 11 states that the maximum usable 
charge sensitivity of the Compton electrometer is 2 X 10 14 
div. /coulomb, which occurs at a voltage sensitivity of 5000 

The Hoffmann electrometer 12 combines the highest charge 
sensitivity of any commercial instrument with stability, 
that is, lack of drift, and ease of working. Great care has 
been exercised to eliminate contact potentials, thermal 
e.m.f.'s, and air currents. To achieve the elimination of 
air currents, heavy copper pieces surround the movable 

10 Pockman, L. T., Rev. Sci. Instruments, 7, 242 (1936). 
ii Wolf, F., Ann. d. Physik, 18, 373 (1933). 
i 2 Hoffmann, G., Phys. Zeits., IS, 480, 1029 (1912). 

Hoffmann electrometers are made by E. Leybold's Nachfolger A. G. 
Koln-Bayental, Bonner Strasse 500, Germany. 


system to insure that thermal gradients are kept at a mini- 
mum. A decided advantage is gained also by evacuating 
the case to a few millimeters of mercury, thus making the 
instrument "deadbeat." 

The instrument operates upon essentially the same prin- 
ciple as the quadrant electrometer. The chief difference is 

bimetal legs 
to maKe temp 
attire variations 
of the su ( , 
equal those 
of the 

detail of inner casing 

Fig. 13. The Hoffmann electrometer. 

that a half vane is used for the movable system, so that, 
instead of quadrants, only two conductors, or binants, are 
necessary. Fig. 13 represents the relationships of the 
essential parts. The platinum needle and mirror together 
weigh approximately 5 mg, and the suspension is a 3/z 
(0.0003-cm) Wollaston wire. 


To achieve a sensitivity independent of temperature, it is 
necessary to keep the vane or needle at the same distance 
from the binants. This is accomplished by inserting into the 
supports of the upper part of the electrometer case, which in 
turn supports the suspension, a metal of such coefficient of 
thermal expansion that the over-all expansion completely 
neutralizes the change of length of the torsion fiber with 

Contact and thermal electromotive forces are kept at a 
minimum by making everything from or plating it with 
platinum. Also, insulation is protected by metal, so that 
possible spurious charges cannot affect the system. 

Since the Hoffmann electrometer combines so many de- 
sirable features, it may be well to list some of them. These 
characteristics must be combined in any other instrument 
with which it is intended to push the charge sensitivity to 
that limit set by Brownian motion, and still have freedom 
from drift and a reasonable working period. 

1. The moment of inertia of the moving system must be 
small. (See Eq. 13.) 

2. The suspension must be made of material which has a 
small coefficient of internal friction; that is, the needle must 
return to zero after a deflection. 

3. Air currents must be kept at a minimum. This means 
that the moving system must be surrounded with heavy 
copper pieces. The suspension should be closely surrounded 
by metal pieces as well. 

4. The case must be evacuated to keep the working time 
within a reasonable limit. 

5. Temperature compensation is needed if the distance 
between the vane and the stationary parts is to remain 

6. Thermal and contact electromotive forces must be 

In addition to the above, it is usually desirable to have the 
scale approximately linear. 


Two additional features of the Hoffmann electrometer are 
(1) an electromagnetic grounding switch and (2) an induc- 
tion ring for inducing a charge on the movable system. 

To facilitate making the necessary electrical connections, 
a control mechanism is supplied with the instrument when 
it is purchased. Although not absolutely necessary, the 
control mechanism is a great aid, since the proper connec- 
tions are made and broken at the right time by only one 

The latest model of the Hoffmann electrometer 13 combines 
all the desirable features of the earlier models but permits 
greater accessibility to the essential parts. Also the ad- 
justments are much more easily made; for example, in the 
older types the instrument had to be exhausted after adjust- 
ing the binants, while in the new design this adjustment is 
made through a sylphon from 
the outside. 

String electrometers are 
divided into two main divi- 
sions: (1) those with a fiber 
supported only at one end 


15/i gilded 
quartz fiber 
4 cm long 

and (2) those in which 
fiber is kept taut by a 
spring. The latter are 
most common and, as far as is 
known, are the only ones on 
the market. 

Electrometers of the first 
class are easily made and are 
often very satisfactory when 
high sensitivity is not needed. 
The two plates can be flat and the fiber hung down between 
them as shown in Fig. 14. There should be an adjustment 
either on the plates or on the fiber or on both to bring the 
mechanical and electrical zeros together. Some adjustment 
may be made by tipping the instrument in the appropriate 

13 Zipprich, B., Phys. Zeits., 37, 35 (1936). 

Fig. 14. An easily constructed 
quartz-fiber electrometer in which 
the fiber is supported only at one 


direction. A microscope must be provided to read the de- 
flection. With the plates 1 cm apart, the fiber should be 

about 25/x in diameter and 4 

cm long if the potential on 
the plates is not to exceed 
100 volts. 

Of the second type, that de- 
signed by Wiilf 14 is, perhaps, 
typical. It is shown diagram- 
matically in Fig. 15. The fi- 
ber is usually a Wollaston wire 
2fx in diameter, kept taut by a 
quartz-fiber bow. Screw ad- 
justments permit movement 
of the plates with respect to 
the fiber as well as change in 
the tension of the fiber. 



wire or 







Fig. 15. Schematic diagram of the 
Wiilf string electrometer. 

cjilded quartz 

jit^ t 

With all string electrometers the deflection is not a linear 
function of the applied charge or voltage at high sensitivity. 
It frequently happens that at high sensitivities the fiber 
leaves the field of view of the 
microscope as it reaches a posi- 
tion where instability occurs. 

The chief advantages of 
string electrometers are (1) 
portability, (2) ease of adjust- 
ment, and (3) short working 

The Perucca electrometer 15 
is similar to the string elec- 
trometer, except that the part 
between the plates consists of 


gilded quartz 

C~^-< quartz bow 

©"^torsion adjustment 

Fig. 16. Principle of operation of 
the Perucca electrometer 

14 Wiilf, Th., Phys. Zeits., 15, 250, 611 (1914). 

This instrument is made by E. Leybold's Nachfolger A. G. Koln-BayentaL 
Bonner Strasse 500, Germany. 

15 Perucca, E., Zeits. f. Instrumentenk., 47, 524 (1927). 

This instrument is obtainable from E. Leybold's Nachfolger A. G. Koln- 
Bayental, Bonner Strasse 500, Germany. 


two conducting quartz fibers supported on a torsion fiber as 
shown in Fig. 16. The two movable fibers are brought 
together at one end, and a small index, which is viewed with 
a microscope, is provided. The charge and voltage sensi- 
tivities are each greater than can be obtained with a string 
electrometer. Not only is it more sensitive, but it also 
combines all the advantages of the latter instrument. 

The Lindemann electrometer 16 was developed primarily 
for use with photoelectric cells mounted on telescopes in 
measuring light from stars. Such use requires that the sensi- 
tivity and position of the moving system be independent of 

quad rants about 
1 crn 


motion of the tip 
of the needle 
in a micro- 





20yu gilded 

glass fibers 
on either 
side of the 
fiber with 
electro lyti- 
tally depos- 
itee! copper 

6yu quartz 
torsion fiber 
ji 1.4cm long; 

detail of suspension 

The whole instrument is enclosed in a metal box 4.5cm, 
by Z. 8cm, by 3cm high, with windows top and bottom. 

Fig. 17. Arrangement of quadrants and movable system in the Lindemann 


tilt. The first is accomplished by making all parts very 
rigid and the second by using for the moving system a light 
needle mounted on a stretched torsion fiber. Since a mirror 
and scale would be cumbersome for such uses, the deflection 
of the needle is read with a microscope with a micrometer 
ocular. The whole electrometer weighs but 80 g. The 
quadrants and needle mountings are represented in Fig. 17. 

16 Lindemann, F. A. and A. F., and Kerley, T. C, Phil. Mag., 47, 577 

This instrument is obtainable from the Cambridge Scientific Instrument 
Company. Ltd.. Cambridge. England. 


The principle on which the instrument works is similar to 
the quadrant electrometer. The quadrants are 1.5 cm broad 
and 1 cm high, with a slot 2 mm wide, into which the needle 
may pass, cut in each. These plates are mounted 5 mm 
apart on quartz rods. The torsion-fiber mounting and 
needle are placed between the plates so that the junction of 
the needle and torsion fiber is symmetrically located with 
respect to the four plates. The whole is mounted in an 
aluminum box with suitable connections. Through a glass 
window in one side of the box the motion of the end 
of the needle is observed with a microscope. A window 
directly opposite on the other side of the case permits fight 
to enter. 

The needle may be balanced so that a rotation of the 
instrument through 90° makes less than 0.06 mm motion of 
the end. The needle is usually about 1 cm long, and the 
torsion fiber is 6/* in diameter. The needle and torsion fiber 
are covered with a conducting coating of metal, and a suitable 
connection is made to the outside of the case. The quartz 
frame for holding the torsion fiber serves as insulation as 

Electrical connections are the same as for any quadrant 
electrometer. Instability occurs at a potential of about 
100 volts. At 3 volts below this unstable value the deflec- 
tion will reach 99 per cent of its final value in 1 second. 
The voltage sensitivity under these conditions is 0.76 mm/ volt 
motion of the end of the needle. With a suitable microscope 
a workable sensitivity of 500 div./volt can be obtained. 
The electrostatic capacity is about 2 cm. 

The instrument may conveniently be used with a leak to 
measure currents of from 10 -10 to 10 -14 ampere. When not 
too great demands are to be met, this quite inexpensive 
electrometer will meet many needs, especially where porta- 
bility is a requirement. 

For a discussion of circuits, sensitivities, and limitations of 
the vacuum-tube electrometer, which uses specially con- 
structed vacuum tubes, see Chapter X. 


Some practical considerations in the use of electrometers 
and electroscopes. Useful sensitivity in X-ray work. 
Electrometers are frequently used with an ion chamber in 
X-ray work. As is well known, ions are formed not only by 
the X-ray beam but by (1) cosmic rays, (2) local radiation 
from radioactive matter in the surroundings, and (3) radio- 
active contamination on the inner walls of the ion chamber. 
Of these, (3) can be reduced to a small value in comparison 
with the others by two effective means. The inner walls 
can be painted with a mixture of collodion and lampblack, 
each of which is quite free from radioactive materials. 
The thickness should be about 0.05 mm to stop all the 
a particles. The other method is to maintain a fine wire 
grid at a suitable potential to drive the ions formed by 
the a particles back into the walls. Since in general (3) is 
due to particles which will have a range of less than 5 cm at 
normal air pressure, the range can be kept within the grid 
by a gas of high molecular weight or increased pressure, 
or both. Even rubbing Carborundum paper on the walls 
of the chamber will often help considerably in lowering the 
emission. As for (1), cosmic rays could be reduced to an 
extremely low value by going into a mine 100 to 200 feet 
below the surface of the ground, while (2) could be made 
negligible with 4 inches of surrounding lead. However, 
since it is not practical to go to such trouble, in most cases 
it is necessary to make the best of the situation. 

Since the error of a result which depends on the difference 
or sum of two readings is 17 

6 = ( €l 2 + 62 2 ) 1 ' 2 , (16) 

where d and € 2 are the errors of the two readings, it is hope- 
less to try to push the sensitivity of the measuring device 
beyond a certain point, and the only hope of increasing 
accuracy is by a longer period of observation. It is easily 
seen that the optimum useful value of the sensitivity is 

17 For a further discussion of probabilities and errors involved when a meas- 
urement depends on the effects of a finite number of particles, see page 298. 


reached when the deflection due to background only and the 
deflection due to the X-ray beam only are equal for the same 
time of observation. 

If the ratio of background to beam readings is to be made 
as small as possible, it is obvious that the volume also should 
be made as small as possible, since the background reading 
goes up with the volume. 

Example. Assume an ion chamber of 1000 cm 3 . Cosmic 
rays will contribute about 3 ions/cm 3 /sec. /atmosphere of 
air at sea level. Local radiation will be from 3 to 5 J (ions 
cm -3 sec. -1 atmosphere -1 of air), while the background may 
vary over wide limits; from 0.1 to 10 ions cm -3 sec. -1 atmos- 
phere -1 of air will probably include the extreme cases. 
Since the a-particle paths will usually end in the gas, an 
increase of pressure will not change the number of ions 
formed by the a particles, but the ionization due to local 
radiation and cosmic rays will go up as the pressure increases. 
Let us assume 77 as due to electrons and 57 as due to a par- 
ticles. The average path length of the electrons is about 
10 cm, and at 60 ions/cm of path this corresponds to about 
12 electrons/sec. crossing the chamber. In order to have a 
mean relative error of ei, then according to the laws of proba- 
bility, 1/ei 2 particles must cross the chamber. Let €i be 
0.03, or 3 per cent; then 10 3 particles must be counted. This 
will take 80 seconds on the above assumptions. These 10 3 
electrons will form 6 X 10 5 ions. Hence if we assume the 
same fluctuations in the ions from the beam, then according 
to Eq. 16, to have an average error of 4 per cent in a reading, 
we must time for at least a minute, and the sensitivity of the 
electrometer need not be greater than 10 -5 div./ion if we 
estimate to 0.1 div. 

As for the a particles from the walls, their effect may be 
considered as follows: Supposing they amount to 57, which 
is not an uncommon value, then there will be 5000 ions/sec. 
formed. Now an a particle will form, on the average, about 
10,000 ions in the gas. Hence there is 0.5 a particle/sec. 
emitted by the walls. Now if €i is the mean absolute error 


in a given reading and e 2 in another reading, the mean rela- 
tive error e r in the sum will be 

_ fe 2 + 62 2 ) 1 ' 2 (±7) 

€r - ~KTer 9 (17) 

where 6 is the deflection of the instrument. If Ni is the 
number of particles per second of one kind of particle and 
7i the number of ions formed per particle, then it can be 
shown from Eq. 17 that the mean relative error of the sum is 

/ NJS + N 2 h 2 Y /2 / (4.3 + 50) X 10 6 \ 
" \t(NJi + Nzh) 2 ) V(0.72 + 0.50) 2 X 10V 

(4.3 + 50) X 10 6 \ 1 / 2 06 


under the above assumptions if the subscript 1 refers to the 
electrons and 2 refers to the a particles. It is then necessary 
to count for 400 seconds to gain an accuracy of 4 per cent. 
In this time 5 X 10 6 ions will have been collected. In order 
to read this to 4 per cent we need a sensitivity no greater 
than 10 -6 div./ion. 

The above calculations have been made to show (1) the 
importance of eliminating a particles as much as possible 
and (2) that when this is done completely, the sensitivity 
of the electrometer has a limit beyond which there is no 

If charges are to be collected where there is very little 
background, such as in photoelectric work, then there is no 
reason why the sensitivity cannot be pushed to the maxi- 
mum. In all cases, if possible, an electrometer or an 
electroscope suitable to the accuracy required should be 

Useful sensitivity in cosmic-ray work. In case the instru- 
ment is subject only to cosmic rays and no shielding is used, 
the ionization is due to random electrons and "X" particles, 
which ionize the gas the same as electrons. When this is so, 
the mean relative error is N~ 1/2 , where N is the total number 
of particles, the effects of which are measured. If there are 
n high-energy particles/cm 2 /sec, and if the mean relative 

error is e r for one reading, then we must observe for a time 


t = 

e r 2 irnR 2 

where — 2 = N f 

for a spherical ionization chamber, and if a is the specific 
ionization, since the average path length is %R, the total 
number of ions collected in the time will be 

4 <jR 

V = 


and this must give a deflection which can be read with no 
larger relative error than e r . If c = 60 ions cm -1 , R = 
10 cm, and e r = 0.01, then v = 8 X 10 6 ions, and we need 
a sensitivity of about 3 X 10~ 6 div./ion. Now n = 0.02 
electron/cm 2 /sec. at sea level. Hence the minimum time 
of observation should be 30 minutes, and for each observa- 
tion the mean error will be 1 per cent. This calculation, of 
course, neglects the error introduced by the background 

Frequently, however, the ionization chamber is surrounded 
with shields made of iron or lead. These do two things: 

(1) In general, they lessen the intensity of the radiation, and 

(2) they introduce new radiations. All the particles passing 
through the ionization chamber are no longer randomly 
distributed in time, for, in addition, there now exist showers, 
consisting of from two to several hundred electrons, which 
come all at the same time from some region of the shield. 
These introduce larger fluctuations than would otherwise 
exist, and the time of observation for the size of ion chamber 
assumed above may be from two to four times as long for 
the same error, and a correspondingly less sensitivity of the 
measuring instrument will serve the purpose. 

Steady deflection measurements. In some cases it may 
be desired to use the constant deflection instead of the drift 
method. This may be done by using the electrometer to 


measure the drop in potential across a fixed resistance as 
shown in Fig. 18. Assume that it is desired to measure a 
constant ion source /. Let the capacity to ground of the 
external system be C\ and of the electrometer c 2 , and let the 
drop in potential be measured across R\. Then 

i\ + k = I', i\ 


, *2 



The equation for the potential across the electrometer is then 

V = [/ - ( Cl + c 2 )f >, 

Solving and putting in the boundary condition that when 
t = 0, V = 0, 

V = IRi\ 1 — e~ (d+c2)Ri \. 

Thus the potential across the electrometer rises exponen- 
tially. If we say arbitrarily that we shall wait until the 
deflection is 99 per cent of the ul- 
timate deflection, then we must 
wait a time t = 4.6 RiC, where 
c = Ci + c 2 . The deflection will 





be approximately VS V after this 
time. Had we measured I by the 
drift method, we should have the 
same deflection in a time RiC, the 
difference, of course, being due to 
the fact that in the second case 
the drift is constant, while in the 
first case the drift begins at the 
same rate, that is, as if Ri = oo f 
but gradually slows down, becom- 
ing very slow toward the last. 

It is therefore much more satisfactory to use the drift 
method for measuring feeble currents, while larger currents 

Fig. 18. Measuring an ion 
current / by determining the 
potential drop across a re- 
sistance i?t by means of an 
electrometer, c 2 . 


are conveniently measured by the steady deflection method. 
The drift method can be used in measuring large currents 
also by inserting a capacity of the appropriate value to 
lengthen the time of drift. 

Limitations of various types of instruments. Limitation 
on the charge sensitivity of electroscopes and electrometers 
has already been pointed out. For the former the maximum 
charge sensitivity is 

(S a W -!<«*)-*» = ^ 

and for the latter 

OWmo. =itac&)- 12 - 

2 V — ' W b 

The capacity of an electroscope which has no external lead 
will depend on the particular design, but for the Wlilf or 
torsion type it will lie between 0.4 and 1 cm. That of an 
electrometer with its added external capacity will probably 
be between 20 and 100 cm. The restoring constant, k, of 
the suspension can be reduced in each to a point where the 
sluggishness of the motion makes the instrument tedious 
to work with, or in the case of most electroscopes, where 
the collecting potential becomes too small to collect most of 
the ions. Since an electrometer case can be evacuated, it 
is possible to adjust the pressure until the motion of the 
vane or needle becomes critically damped. 

If the electrometer case is not evacuated, the working 
period may become excessively long when high sensitivities 
are desired. Much can be achieved by making the needle 
or vane small and light, as is done in the Lindemann and 
Perucca electrometers, and as is inherently the case with 
string electrometers. 

Limitations imposed by drift. The amount of drift during 
a reading is often the limiting factor in electrometers. This 
frequently becomes bothersome long before the maximum 
sensitivity has been reached. One of the chief reasons for 
the drift is that the mechanical and electrical zeros gradually 


drift apart. The deflection caused by the zeros being differ- 
ent may be many times the actual amount they are apart. 
Drift, among other causes, is due to (1) fluctuations in 
battery voltage and (2) nonelastic changes of strain in the 
suspension. If the drift were constant, proper allowances 
could be made, but there are so many factors which depend 
in a different way upon changes of voltages, temperature, 
humidity, and so forth, that it is often very difficult, if not 
impossible, to eliminate completely or take account of the 
drift. This is especially true with vacuum-tube electrom- 
eters, even though balanced circuits are used. 

Limitations on the amount of useful magnification. Two 
methods are in general use for determining the amount of 
deflection in an electroscope or electrometer: (1) Microscope 
with micrometer ocular and (2) mirror and scale. For the 
Lindemann electrometer and most electroscopes the micro- 
scope is used. The limitation as far as magnification is con- 
cerned amounts to a limitation of resolution. Magnification 
can continue until the position of a diffraction band cannot 
be located to within 0.1 div. in the eyepiece. Beyond this, 
nothing is gained. With a numerical aperture of 1 and an 
image distance of 20 cm the shortest useful focal length is 
about 3 mm with a 100-div. scale in the eyepiece 1 cm long. 

If a mirror and scale are used, there is a certain minimum 
mirror size which will allow sufficient resolution. With a 
1 mm div. scale at the customary distance of 1 m, it is 
necessary to have a mirror at least 2 mm in diameter to read 
to 0.1 div. on the scale. 

In all cases, whether in resolution, amount of drift, fluctua- 
tions, or the like, it should always be possible to estimate to 
0.1 of the smallest division on the scale, and in general it is 
useless to push the sensitivity of any instrument beyond the 
point where 0.1 div. loses its significance. 

Limitations imposed by Brownian motion. It is part of 
the classical theory of the equipartition of energy that all 
bodies have a mean thermal energy of %KT for each degree 
of freedom, where K is Boltzmann's constant and T the 


absolute temperature. This Brownian motion of the in- 
strument is evidenced by random fluctuations about the 
point of equilibrium. It is evident that before a super- 
imposed steady deflection can be detected, it must be at 
least as large as this mean Brownian deflection. 

The mechanical energy of a moving system with a restoring 
force proportional to the displacement is J&J 2 , where k is 
the restoring force (or torque) per unit of displacement. 
If A£ is the mean Brownian deflection, then the k corre- 
sponding to this is given by 


_ KT 

~ (A£) 2 ' 

Now with the electroscope the maximum charge sensitivity 
is reached when S Q = %(ck)~ l/2 . Consequently, the corre- 
sponding charge sensitivity is 


So = 



if the deflection is equal to the mean Brownian deflection. 
At room temperature the maximum charge sensitivity is 
thus limited for electroscopes to 

: Q , 1.2 X 10- 4 ,. , . , 

(oo)max. = i72 div./electron, 


where c is in centimeters. For electrometers the expression 

becomes _ _ A 

, Q * 0.8 X 10- 4 /. 71 , 

(ojmax. — Yh div. /electron. 


It is obvious that the electrostatic capacity of the instru- 
ment should be as small as possible if it is intended to push 
the charge sensitivity to the limit. Inherently the capacity 
of the electroscope is much less than that of the electrometer. 
This not only makes it possible to have a higher charge 
sensitivity for the same torsion constant but allows it to be 


It is interesting to compare the above limit with that ob- 
tainable with a Geiger counter. In some applications the 
number of counts and the number of unit charges collected 
are comparable. The mean error with a Geiger counter in a 
single count of N particles distributed at random is N 1/2 , so 
that if it is desired to have a mean relative error of 1 per cent, 
it is necessary to count 1/(0.01) 2 or 10 4 particles. With an 
electroscope having a capacity of 0.5 cm, it is necessary to 
collect 8 X 10 4 electrons to have the same mean error if 
the deflection can be read to 0.1 div. This is, of course, 
disregarding the backgrounds in each case. 

A comparison of various types of instruments. Probably 
the most sensitive electrometer on the market is the Hoff- 
mann. The maximum sensitivity which can be reached 
with this instrument is approximately 5 X 10 15 div. /coulomb. 
Drift has been eliminated to such an extent that sufficient 
time can elapse to detect an average of 1 electron/sec. 
For ease of working, however, it is advisable to keep the 
charge sensitivity in the neighborhood of 1 X 10 15 div. /cou- 
lomb. Much is gained in the Hoffmann by evacuating the 
case, thereby not only shortening the working time but 
greatly eliminating the effects of convection currents. 

The vacuum-tube electrometer has gained much favor in 
the past few years. It has the advantage that it can be 
used in places where it would be inconvenient or impossible 
to use the conventional type of electrometer. The sensi- 
tivity can be made comparable to that of the Hoffmann, 
although it is very much inferior as far as drifts are con- 
cerned. Ordinary precautions consist in having large stor- 
age batteries for plate and filament supply which are kept 
at as constant a temperature as possible, with all leads well 
shielded. Resistances must also be kept constant. Al- 
though with the proper circuit and circuit constants the 
effects of voltage fluctuations are reduced to a minimum, 
it is still not possible to eliminate the drift, and it is usually 
necessary to wait several hours after the connections are 
made for conditions to become only approximately steady. 


When possible, an instrument should be chosen for the 
problem at hand. Frequently it is desirable to use an elec- 
troscope in place of the electrometer. The advantages to 
be gained may be listed as follows: (1) Freedom from ex- 
ternal changes of temperature and humidity, (2) freedom 
from changes in battery potentials and resistances, (3) 
freedom from drifts, (4) need for only one potential, (5) ease 
of setting up and operating, (6) portability, and (7) low cost. 
The disadvantages are that (1) except with the torsion type 
the sensitivity is not as high as with the ordinary electrom- 
eter, (2) the sensitivity is not readily varied, and (3) it is 
not convenient to use a null method of reading. 

In Table I are listed the approximate characteristics of 
some instruments. The values of charge sensitivities listed 
are not the maximum attainable but represent those that 

Comparison of Characteristics of Various Instruments 


S v 

(X 1G 14 ) 

Working Period 

Wulf bifilar 















Wilson tilted 

Neher torsion 


Dolezalek quadrant 






Wulf string 




Hoffmann vacuum 

Vacuum tube 


Units of voltage sensitivity, Sv, are divisions per volt which correspond to 
the maximum usable charge sensitivity, S q , expressed in divisions per coulomb. 
Values of S q are for no added external capacity. 

can be reached and worked without great difficulty. The 
values of the voltage sensitivities are those which correspond 
to these values of the charge sensitivities. In some cases 
the voltage sensitivity can be made much higher, in particu- 


lar with the Compton, with which it is possible to reach 
50,000 div./volt. The working period represents approxi- 
mately the time for the deflection to become zero after the 
net charge is removed. 

Useful techniques in electroscope and electrometer work. 
Mounting gold leaves. Gold leaf usually comes in sheets 
about 8 cm square, the leaves being separated by sheets of 
tissue paper. The leaf will be found quite uniform and thin 
enough so that objects can be distinguished through it 
when it is held before the eye. The thickness is usually 
about 0.08/x. The leaf is cut to the desired size by placing it 
between sheets of tissue pa- 
per and using a razor blade. 
The paper separating the 
gold leaf will be found satis- 
factory for the purpose. The 
cutting should be done on a 
flat base, such as cardboard. 
If the razor blade is sharp, 
the cut will be clean and the 
gold will not adhere to the 
paper. The leaf can be 
moved around from one sheet 

edge of 

Fig. 19. 

Mounting the leaf on a gold- 
leaf electroscope. 

of paper to another by means of clean needles, mounted so 
that they can be handled with ease. It can also be picked 
up with clean fine-pointed tweezers. If the leaf touches any- 
thing which has a film of organic substance on the surface, it 
will easily adhere with only slight pressure. Once the leaf 
has stuck, it will usually tear before coming loose. It is 
safest to handle it as little as possible. When mounting the 
foil on the single-leaf type of electroscope, it is cut to size 
and then transferred to a piece of paper, such as typing paper, 
and placed so that one end of the leaf is near one edge of the 
paper. The edge of the paper is allowed to overhang the table 
about | inch. Some alcohol-dissolved shellac is spread across 
"&at part of the metal piece from which the leaf is to hang. 
The edge of this shellac must be perpendicular to the edge of 


the metal support, in order that the leaf, when mounted, will 
deflect in a plane perpendicular to the plane of this support. 
The metal piece must be clean, or the leaf is apt to adhere 
to it. The metal support is brought into position as shown 
in Fig. 19 and then lowered gradually. The paper will 
bend and the leaf will adhere to the shellac. 

The above operations should be carried out in a room in 
which the motion of the air is at a minimum. It is often 
advisable to wear a mask or deflector over the nose to avoid 
blowing the leaf about. 

Preparation of Wollaston wire. 18 The Hoffmann and many 
string electrometers use a fine platinum suspension known 
as Wollaston wire. It may be obtained in various sizes 
from 1.5/x to 5/z. To produce such a fine wire of uniform size, 
the following process of manufacture is used. Upon a much 
larger platinum wire is electroplated a uniform layer of 
silver. The combination is then drawn down until the fine 
thread of platinum in the middle is of the proper size. The 
silver is etched off with acid. Since the resulting platinum 
wire is quite delicate, special care must be used in the etching 
as well as in subsequent handling. 

In order to avoid small bubbles collecting on the wire and 
interfering with the etching, or in some cases breaking the 
fine wire, a special solution of chemically pure nitric acid in 
distilled water at a density of 1.10 g cm -3 is used. To insure 
uniform etching, the wire should be thoroughly cleaned 
before it is immersed in the acid. As an aid to handling after 
etching, a bead two to three times the diameter of the silver 
wire is formed on one end with a small oxygen flame before 
the silver is etched off. A section of wire is then cut off, 
perhaps an inch longer than the necessary suspension. The 
solution is placed in a tall vessel, such as a graduate, and the 
straightened silver wire is supported in it vertically. The 
suspension should be left in for a longer rather than a shorter 
time, since the platinum is not damaged by the solution. 

18 Wollaston wire is obtainable from Hartmann and Braun, A. G., Frank- 
furt am Main, Germany, and Baker and Company, Philadelphia. 


It is necessary that all the silver be etched off, or the sus- 
pension may be ruined in the annealing process. The small 
bead marks the lower end as the suspension is drawn from 
the solution. Before soldering it into place, it may be de- 
sirable to mount the suspension on a " wishbone," the dis- 
tance between the two prongs being somewhat greater than 
the length of the mounted suspension. If quartz fibers 20/x 
to 30/jl in diameter are mounted in the tips of the prongs and 
the Wollaston wire is fastened to these with a hard wax, 
there will be much less chance of breakage. 

Either before or after mounting, the suspension should be 
placed in a horizontal position and annealed with a small gas 
flame. In still air the flame is passed beneath at such a 
distance that the platinum is heated to a bright red color. 
If all the silver has been etched off, the suspension will 
appear a uniform brightness throughout its length. The 
annealing is necessary, if the wire is to be used in an elec- 
trometer, to relieve the strains which resulted from the 

In soldering the suspension in place, a c.p. solution of 
zinc chloride is a good flux. The heat is best applied with a 
small soldering iron, not directly at the point at which the 
suspension touches the solder but at a short distance away, 
relying upon the conductivity of the metal support. It is 
best to work under a magnifying glass or, better still, a 
binocular microscope. The joint should be rigidly inspected 
to see that the platinum is actually embedded in the solder 
and not just held by the solidified flux. 

Insulators used in electrometer and electroscope work. 
The insulator ordinarily used in electrometers is amber. 
The amber now on the market is usually a manufactured 
product which has as good insulation properties as the natu- 
ral amber and has the advantage of being obtainable in a 
variety of sizes. Amber has a high volume resistivity, and 
the surface resistance of clean amber is also high. If the 
surface is contaminated, the best remedy is to remove some 
of the amber with a clean tool by turning it in a lathe. If 


this is not convenient, the amber may be covered with a thin 
coat of ceresin, as will be described later. 

The best insulator known is clean, dry, fused quartz. 
By clean quartz is meant quartz which has not touched 
anything since being heated to the softening point, and by 
dry quartz is meant quartz either in a good vacuum or in a 
gas dried by phosphorus pentoxide. Fused quartz is also 
superior to other insulators in that the soak-in is far less. 
Under comparable conditions amber has at least ten times 
the soak-in possessed by quartz. 

Ceresin is a natural wax which has remarkable electrical 
insulation properties. 19 It is about the same hardness as 
ordinary paraffin, each at 20°C. However, it has a some- 
what higher melting point than either paraffin or the arti- 
ficial ceresin, being liquid at 65 °C. Its insulation properties 
have been measured by Curtiss 20 of the Bureau of Standards. 
He gives the surface resistivity as greater than 10 17 ohm cm 
even at 90 per cent humidity. One of its main uses in the 
laboratory is to improve the surface resistance of other in- 
sulators. If the solid insulator and the ceresin are each 
heated to around 100°C. and a light coating of ceresin applied, 
the surface leakage will usually be found greatly reduced, 
sometimes by a factor of 100. 

19 Natural ceresin is distilled from the mineral ozokorite. An artificial 
ceresin, which is inferior to the natural product, is also on the market. In 
ordering, the natural product should be specified. 

20 Curtiss, L. F., Bulletin of the Bureau of Standards, 1915. 


Geiger Counters 


H. V. Neher 

string electrometer 

PT^HE Geiger counter is an ion-magnifying device which is 
-*- sensitive to individual ionizing particles. The resultant 
flow of charge, except for the so-called proportional counter, 
is practically independent of the number of ions formed by 
the original particle. Thus, in most Geiger counters an 
a particle forms from 10 2 to 10 3 times as many initial ions 
as a p particle; yet each gives rise to a pulse of nearly the 
same size, and each is usually registered as one particle. 

These counters have now reached a practical state of high 
development as a means of studying feeble radiations, such 
as those found in cosmic rays 
and artificial or natural radio- 
activity. The mechanism of 
the gaseous discharge in the 
counters is, however, not well 
known. ' 

The point counter. The 
original design of Geiger 1 
consisted of a pointed wire 
surrounded by, and insulated 
from, a metal cylinder as 
shown diagrammatically in Fig. 1. A high potential of 1500 
to 5000 volts is applied across the counter, through a high 
resistance R (about 10 9 ohms). The cylinder is made 

1 Geiger, H., Verh. d. D. Phys. Ges., 15, 534 (1913); Phys. Zeits. t U. 1129 


Fig. 1. The original point counter 
and circuit used by Geiger. 



[Chap. VII 

positive with respect to the point P. The pulse is observed 
by means of a string electrometer. 

Briefly, the action of the counter may be described as 
follows: The electric field immediately around the point is 
high enough so that at the pressure of the gas used any ion 
entering the space builds up by collision a large number of 
ions, which in turn build up more ions, until the quantity 
of charge which finally flows reaches the order of 10~ 8 
coulomb, depending, of course, among other quantities, 
upon the applied potential. This charge, collecting on the 
distributed capacity, which may be represented by C, 
causes the potential across the counter to drop to a point 
at which the discharge can no longer be maintained, and the 
charge leaks off across the resistance R. The circuit then 
returns to its normally sensitive condition and is ready for a 
second count. The charge which builds up on C causes a 
drop in potential across R which is read by the electrometer. 
The proportional counter. The counter just discussed 
operates on a trigger principle, and the size of the pulse is 

practically independent of 
the ionization of the initial 
particle, responding alike to 
a or j8 particles. However, 
Geiger and Klemperer 2 have 
found that if a small metal 
sphere is fastened to the 
point and made positive with 

A B* C 


Fig. 2. Characteristic curve for 
the proportional Geiger counter. 
In the region from A to B the size of 
the pulse is proportional to the 
original ionization. In the region 
C to D the size of the pulse is prac- 
tically independent of the amount 
of initial ionization. 

respect to the cylinder, in- 
stead of negative as in 
Geiger's original design, over 
a limited range of voltage, 
that is, within the range A 

2 Geiger, H., and Klemperer, O., Zeits. f. Physik, 49, 753 (1928). See also 
the following: 

Franz, H., Zeits. f. Physik, 63, 370 (1930). 

Klarmann, H., Zeits. f. Physik, 87, 411 (1934). 

Duncanson, W. E., and Miller, H., Roy. Soc., Proc, A, U6, 396 (1934). 

Haxel, O., Phys. Zeits.. 36, 804 (1935). 

Chap. VII] 



to B in Fig. 2, the pulse is approximately proportional to the 
original ionization of the particle. This circumstance makes 
it possible to distinguish between heavy particles, such as 
a rays, protons, and deuterons, and the much lighter parti- 
cles, electrons. Since y rays show themselves by the elec- 
trons liberated from the material they pass through, it is 
also possible to count individual heavy particles in the pres- 
ence of strong X-ray and y radiation. 

A design of proportional counter which has been used by 
a number of workers is shown in Fig. 3. The cylinder is 

t-to 2-mm steel ball soldered 
to # 2.4 brass wire 





inches t 

I ' ■ ' ' i ■ ■ ■ v / 

O 1 2 





Fig. 3. Typical construction of a proportional counter. 

maintained at a constant negative voltage with respect to 
the ball and wire. A steel ball from 1 to 2 mm in diameter 
will give good results, but the metal of which it is made is 
unimportant. The wire or rod supporting the ball may be 
made of almost any convenient metal, and is usually about 
one half the diameter of the ball. When a heavily ionizing 
particle enters the sensitive region surrounding the ball, 
negative charges are collected. The effect is amplified by 
a linear amplifier. Since the amount of charge collected is 


proportional to the initial ionization, a means is here provided 
of distinguishing between heavy particles and electrons. A 
thin cellophane, lacquer, or mica window over the opening 
permits suitable gases to be used at reduced pressures. 
The threshold potential will probably lie between 1500 and 
5000 volts, depending on the kind of gas used, its pressure, 
and the geometry of the counter. 

Brubaker and Pollard 3 have studied the effects of various 
gases, using different kinds of heavy particles. Their recom- 
mendations are to use argon at pressures greater than 50 cm 
of mercury if there is a background of y radiation, while for 
a-particle or proton-scattering experiments, where y rays 
are not serious, hydrogen, nitrogen, or air between 2 and 
10 cm of mercury pressure can be used. 

If the potential on the above counter, with the small 
sphere in place of the point, is raised, a region C to D in 
Fig. 2 is reached where the effect of all particles is practically 
the same and the number of counts per unit time becomes 
almost independent of the applied voltage for a constant 
radiation. The length of this "plateau" depends primarily 
upon the distributed capacity and the resistance across the 
counter. When the point D is reached, the number of counts 
for a given radiation begins to increase, and any further in- 
crease in voltage soon sets up a steady gaseous discharge. 

Since the sensitive portion in a point counter is limited 
to a small region in the immediate neighborhood of the point, 
it is useful primarily in experiments in which only a small 
solid angle is to be studied at a time, such as in problems on 

The "Zahlrohr" or Geiger-Mtiller counter. When it be- 
comes necessary to have a large area sensitive to ionizing 
particles, the point counter no longer can be used, and its 
place is taken by the Zahlrohr or Geiger-Mtiller 4 counter 
(hereafter designated as the G-M counter). It has become 
particularly useful in the study of cosmic rays, for it is possi- 

3 Brubaker, G., and Pollard, E., Rev. Sci. Instruments. 8. 254 (1937). 

4 Geiger, H., and Muller, W. t Phys. Zeits., 29, 839 (1928). 

Chap. VII] 



C=30cm T\i=lo 9 ohms I^=£xio T ohrt)S 

Fig. 4. A method for studying 
qualitatively the action of a Geiger- 
Muller counter, using a string elec- 
trometer as the detector. 

ble to use tubes of large cross-sectional area and thus have 
an accuracy comparable to that obtained with ionization 
chambers for the same time of observation. The advantages 
to be gained over the ionization-chamber method are that, by 
properly combining two or 
more G-M counters, (1) par- 
ticles incident only from lim- 
ited angles can be counted, 
and (2) background radia- 
tion due to contamination on 
the counter walls and radio- 
activity of the surroundings 
may be eliminated. 

Behavior of the G-M counter. 
Some of the properties of the 
counter may be studied by 
the arrangement shown in 
Fig. 4. The string electrometer, E, will be suitable, pro- 
vided the counter is small, that is, the number of counts 
per minute is not more than 10 to 20. The circuit constants 
should be approximately as follows: Ri = 10 9 ohms, C = 

30 cm, and R 2 '= 20 X 10 6 
ohms. The tube is made nega- 
tive with respect to the wire. 
As illustrated in Fig. 5, when 
the potential, V, is raised, no 
effect will be noticed on the 
electrometer until a certain po- 
tential, known as the " thresh- 
old" voltage, V t> is reached, 
when a small increase in potential will cause the number of 
counts per unit time to rise quite abruptly to a certain 
value. 5 Any further increase in the potential will cause 

6 The abruptness of this rise depends, to a large extent, on the ratio of the 
length to the diameter of the metal tube and on the position of the central 
wire. Owing to the end effects, the part of the tube which first becomes sensi- 
tive is that near the center. As the potential is raised, this active region moves', 
out toward the ends, and soon practically the whole length becomes sensitive. 


voltage V t v v rr 

5. Characteristic curve 
Geiger-Muller counter. 

of a 


very little change in the counting until a potential V m , which 
may be called the maximum operating potential of the 
counter, is reached. A small increase from here on will 
cause a sudden increase in counts, which soon goes over into 
a glow discharge. The "plateau" for a good counter and 
proper circuit constants may be 200 to 300 volts or even 
longer. This fact permits quantitative results to be ob- 
tained with G-M counters without elaborate means of regu- 
lating the voltage supply. The tube is usually operated at 
some intermediate voltage, V. 

Although both positive and negative particles are present 
in the tube, the actual multiplying agents are probably the 
electrons. The electric field is higher than necessary for 
the electrons to form ions by collision, while it is probably 
not high enough for the positive or negative ions to do so. 
The electrons rushing toward the wire form new positive 
ions and electrons, the current building up according to the 
law i = i e ax , where a is the number of new pairs of ions 
formed per centimeter of path and is called the Townsend 
coefficient. Probably negative ions are also formed by the at- 
- tachment of electrons to the 

\ ^ T I molecules. In the ionization 

process light is given off, lib- 
erating new electrons from 
the metal tube, and these, in 
turn, form other ions as they 
rush toward the wire. This 
photoelectric process has 
been found by Christoph and 
Fig. 6. The action of a counter is Hanle 6 and by Locher 7 to £>e 
best studied with a cathode-ray i mpor t a nt in the mechanism 


of the discharge. The process 

of accumulative ionization continues until the potential differ- 
ence between the cylinder and the wire has dropped to a point 
where ionization by collision can no longer occur. The po- 

6 Christoph, W., and Hanle, W., Phys. Zeits., 84, 641 (1933). 
' Locher, G. L., Frank. Inst., J., 216, 553 (1933). 

R^ U= 

Chap. VII] 



tential recovers itself according to the time constant R X C of 
the circuit, C being the distributed capacity as well as the 
capacity of the coupling condenser. 

The best way, however, to study the action of a G-M 
counter is to connect it directly to the two deflecting plates 
of a cathode-ray oscillograph as shown in Fig. 6. The other 
pair of oscillograph plates is connected to a linear sweep cir- 
cuit whose frequency can be varied. If R is about 10 9 ohms, 
then as the potential V is raised, a point is reached at which 
deflections of the electron beam will occur at random inter- 
vals of time, indicating that the G-M tube has started to 
count. As V is raised farther, the average number of pulses 
per unit time remains the same, but the magnitude, as shown 
by the oscillograph, increases by nearly the same amount 
that V increases. By such means it can be shown that the 
potential to which the voltage falls during a discharge is not 
far below the threshold, the actual amount being roughly 
proportional to the difference between the applied voltage 
and the threshold potential. 

The character of the discharge should be that shown in 
Fig. 7(a), where V T is the threshold voltage. Counters 

(si) (b) (c) 

Fig. 7. Three typical discharges of a G-M counter as seen with the oscillo- 
graph, (a) represents the shape of discharge shown by a fast counter; 
(b) and (c) are representative of slow counters. The width of the pulse in 
(a) may be made as short as 10~ 6 second; that for (b) or (c) may be as long as 
0.2 second for a very slow counter. 

may be divided into two main classes, "fast" counters and 
"slow" counters. The drop in potential across the counter 
is extremely rapid for a fast counter, while the recovery time 
depends upon the product of R and the distributed capacity 


of the circuit. If the counter is a slow one, the breakdown 
will be much less rapid, and the potential may remain near 
the threshold for a relatively long time, as much as 0.2 second 
in some cases, as represented in Fig. 7(b). When a radio- 
active source is brought up, the time spent by the counter in 
a continuous discharge state near the threshold may in- 
crease, so that the counter in a recording circuit would appear 
to be insensitive to radioactivity or even to have a negative 
sensitivity. In some cases the shape of the discharge is that 
shown in Fig. 7(c), where the breakdown is rapid at first 
but the counter fails to recover immediately, the potential 
fluctuating over wide ranges until recovery finally sets in. 

Fast counters will retain the shape of discharge curve • 
shown in Fig. 7(a) when the resistance R is decreased to as 
low as 10 5 ohms. The length of the pulse in this case will be 
about 10~ 5 second. The best counters will still extinguish 
themselves when R is made as low as 4000 ohms. The width 
of the pulse on the oscillograph in this case cannot easily be 
measured, but it should be less than 10~ 6 second if C is 
about 25 micro-microfarads. 

Counters with this short time constant have important 
applications when high counting rates are to be measured or 
when the number of accidentals in a coincidence circuit is to 
be kept at a minimum. In fact, it is usually necessary to use 
only two G-M counters in a coincidence circuit, for, as will 
be shown later, if each of the two counters counts on the 
average 3 times a second, then with a pulse of time width 10 -5 
second there will be an average of only 16 accidentals per 

A complete explanation of the action of these counters 
cannot be given, but it appears that the chief agents causing 
the pulse are electrons and not negative ions, since the latter 
have much too low a mobility to be collected in these short 
times. In a slow counter there is a delaying action of 
some sort, and charges are collected over a relatively long 
period of time. It seems probable that negative ions as well 
as electrons are collected at the central wire in this case. 

Chap. VII] 



As will be discussed presently, the surface of the cylinder has 
a vital effect upon the action of the counter. Among the 
possible physical properties which might be altered by treat- 
ment and affect the counter action could be mentioned (1) 
the work function of the surface, which would affect its 
photoelectric properties, and (2) the electrical resistance of 
the surface of the metal cylinder. At the present time, as 
already mentioned, very little is known of counter action. 

Something can be said, however, as to the treatment which 
will give counters the very desirable characteristic of furnish- 
ing an extremely short pulse. This treatment, which will 
be discussed later, is not always necessary. In fact, counters 
with copper cylinders which have never been treated have 
worked well with 10 5 ohms across them. There is, at present, 
no rule by which it can be predicted whether or not a counter 
will have this short reaction time. The proper procedure is 
to try the counter in a circuit such as that shown in Fig. 13. 
If it fails to work, then an oscillograph would show that 
when a potential slightly above the threshold is reached, the 
counter will discharge once and then remain in a continuously 
conducting state. 

Construction of G-M counters. The simplest method of 
making a G-M tube is to take a copper or brass tube of a 

copper tube 

^ tungsten 
^^ wire 

Fig. 8. Simple construction of a G-M counter. 

length at least five times the diameter, insert a hard-rubber 
plug in each end, and pass through the plugs, coaxial with 
the cylinder, a straight wire 0.1 to 0.2 mm in diameter. 
(See Fig. 8.) The wire can be made of many metals, but 
tungsten or copper will give good results. The whole must 
be made tight and the gas pressure reduced to from 3 to 6 cm 


of mercury. The gas may be air or a mixture of air and one 
of the noble gases, particularly argon. 

A counter made in this fashion may work satisfactorily 
for a while, but it is not suited for constant operation over 
long periods of time. Even though the tube could be made 
perfectly tight, which is difficult, it still suffers from the 
defect of having a high temperature coefficient. Curtiss 8 
has shown that with a tube having hard-rubber ends, the 
count goes down as the temperature increases, indicating an 
increase in the density of the gas, probably due to the out- 
gassing of the hard rubber. This temperature coefficient 
can, in practice, be completely eliminated by sealing the 

copper tube *^ 3o mil tungsten wire 

seal off here after evacuating 
and fillingf with air or a 
mixture of air and argfon 

Fig. 9. Typical construction of a copper-in-glass counter. Following this 
general design, counters have been made from 0.5 cm to 10 cm in diameter. 

metal tube inside a glass tube and making metal-glass seals 
to the tungsten wires. The construction details of a counter 
of this latter design are shown in Fig. 9. Satisfactory 
counters of this copper-in-glass type have been made from 
0.5 cm to 10 cm in diameter. 

After the counter is assembled, a concentrated solution of 
nitric acid (12 to 16 normal) should be admitted and allowed 
to attack the copper vigorously for 10 to 20 seconds. The 
acid is then removed and the counter washed thoroughly 
with distilled water. Under this treatment the copper will 
turn a dark, almost black color, which probably is due to a 
thin layer of CuO. The tube is then dried, evacuated, and 
the desired amount of gas admitted. 

8 Curtiss, L. F., Bureau of Standards, J. of Research, 10, 229 (1933). 


The kind and amount of gas used determine to some 
extent the action of the counter. For many purposes air 
admitted to a pressure of 3 to 6 cm of mercury results in a 
very satisfactory counter. A mixture of argon and 10 to 20 
per cent air for the same total pressure will have a threshold 
perhaps 40 per cent lower than air alone. There seems to be 
little choice, however, between the mixture of air and argon 
and air only. The counter will not work with pure argon. 
The threshold potential for a counter, 2.5 cm in diameter 
with a 10-mil wire, filled with 5 cm of mercury pressure of 
argon and 1 cm of air, will be about 800 volts. The same 
counter filled with air to the same total pressure will have a 
threshold of about 1200 volts. 

Counters made according to the above directions will, in 
general, be of the slow type, that is, the collection time for the 
ions will be of the order of 0.1 to 0.01 second. Such counters 
are quite satisfactory for many purposes where short reac- 
tion times are not necessary. They may be used in the 
conventional Geiger circuit shown in Fig. 12, or if it is 
desired to eliminate the high resistance, a radio tube may 
be used to help the counter recover itself as shown in Figs. 14 
and 15. 

In case it is desired to make a fast counter, that is, one in 
which the collection time of the ions is of the order of 10 -5 
second, a different treatment of the copper cylinder is neces- 
sary. The treatment to be described is one of several known 
to produce a fast counter. A counter so treated will have 
the following characteristics: (1) The threshold potential 
will be as low or lower than for the same size counter filled 
with a mixture of argon and air at the same pressure. 
(2) The length of the plateau will be at least 30 per cent of the 
threshold potential. (3) The counter will function in the 
circuit shown in Fig. 13 with only 100,000 ohms in series 
with the high potential, instead of the 10 9 ohms necessary 
for a slow counter. (4) The efficiency is high. By ampli- 
fying the pulses, the efficiency of a 2f-inch counter was 
found to be 100 per cent at a counting rate of 30,000 per 


minute within the limits of experimental error, which may 
be taken to be 1 per cent. 

The procedure to make a fast counter is as follows : 

1. Starting with a copper-in-glass counter with a tungsten 
wire, clean the copper thoroughly with about 6 normal 
nitric acid. (A water aspirator is indispensable for admitting 
and removing solutions.) Such a concentration of acid will 
leave the copper very bright. 

2. After rinsing well, introduce a solution of 0.1 normal 
nitric acid. This will remove any copper compounds 
formed by the stronger acid. 

3. Rinse thoroughly (at least 10 times) with distilled water 
and dry. 

4. With dry air inside, heat the whole counter in a large 
flame until the copper turns a uniform brownish-black color. 

5. Seal the counter off temporarily and then heat for sev- 
eral hours at about 400 °C. Upon cooling, the copper 
cylinder will be coated with the bright red oxide, Cu 2 0. 

6. Evacuate and admit dry N0 2 gas to a pressure of 1 
atmosphere. (This gas can be made by the action of 16 
normal nitric acid on copper. It may be dried by passing 
through CaCl 2 and P 2 5 .) 

7. Heat the counter with the N0 2 until the Cu 2 turns 
a dark velvety color. Pump out the N0 2 . 

8. Admit argon (commercial, 99 per cent pure is satis- 
factory), which has been bubbled through xylene, to a 
pressure of 6 to 10 cm of mercury pressure. The counter 
should be tried at this point. For a 1-inch counter the 
threshold should be 600 to 800 volts for 8 cm of mercury 
pressure. If the counter does not work properly, the gas 
should be pumped out and more argon, which has been 
bubbled through the xylene, admitted. 

9. When the counter is found to work satisfactorily, it 
may be sealed off. 

Although all the above steps may not be necessary in 
all cases, yet this procedure has been found to give verv 


satisfactory counters having reaction times of 10~ 5 second 
or better. The characteristics of the counters also seem to 
be permanent. The photoelectric properties as well as the 
electrical resistance of the surface are probably radically 
changed by this treatment. 

The use of a cathode-ray oscillograph is indispensable in 
the proper study of the action of a counter, and its use 
cannot be too strongly recommended. 

The above treatment is limited to copper-in-glass counters, 
and no method as yet has been found to be applicable to 
counters in general. 

Sensitivity of counters to ionizing particles. If a set of 
three identical counters are arranged one above the other, 
with their axes parallel and hori- 
zontal, and are connected to a cir- /j\ S7\ 
cuit (see page 290) which responds ^\s ^ix 
only to coincidences between the ' i 

three counters, then if the middle Jr^\ 


one is moved out of line, as shown , 

in Fig. 10(a), the counting rate • 

will begin to fall and, except for 

accidentals, which produce a 

small effect, the rate becomes 

zero when the middle one has ' 

been moved far enough so that a Fi £- 10 - Testing the sensi- 

, . . , . tivity of different regions of a 

single particle cannot pass counter, 
through all three. (The count- 
ing, when the three are in a line, is due, of course, to cosmic- 
ray particles passing through the three counters.) Compar- 
ing this rate with the rate due to double coincidences between 
the two outside counters, Street and Woodward 9 have shown 
that for counters 3.82 cm in diameter the effective diameter 
is the same as the geometrical diameter. Similarly, by rotat- 
ing the middle counter 90°, as shown in Fig. 10(b), and mov- 
ing it parallel to its axis, the sensitivity along the counter can 
be obtained. With the conditions under which they were 

9 Street, J. C, and Woodward, R. H., Phys. Rev., 46, 1032 (1934). 


operating and with a counter whose geometrical length was 
13 cm, the effective length was found to be 10.5 cm. The 
difference is probably dependent on the geometry of the 
arrangement as well as the potentials used. 

A set of G-M tubes arranged in a vertical line and con- 
nected to count coincidences provides a means of determin- 
ing the efficiency of a counter. If the middle counter were 
100 per cent efficient, that is, if it responded to all cosmic- 
ray particles passing through the outside counters, then, if 
proper allowance were made for the accidentals, there should 
be as many counts when the middle one is turned off as 
when it is turned on. (The counter should not actually be 
removed, since the amount of absorbing matter will then 
be changed.) By comparing the rates in the two cases, an 
actual measure of the efficiency can be obtained. A good 
counter should be at least 95 per cent efficient. It is im- 
portant that the efficiency for counters in a coincidence 
circuit be high, or the number of counts will be greatly re- 
duced. If the efficiency of each of n counters is e, the 
number of coincidences will be only e n of the number which 
would be counted if they were 100 per cent efficient. 

G-M tubes for special uses. The tube illustrated in Fig. 9 
answers very well the needs for work on cosmic rays, since 
the radiation readily penetrates the thin walls. For other 
types of radiation special constructions are necessary or are 
more efficient. A few of these will be discussed briefly. 

For /3-ray measurements, a point counter, as discussed 
on page 259, can be used at atmospheric pressure, in which 
case no trouble is experienced by the particles in entering 
the counter. In case the cylindrical counter is preferred, 
there are two alternatives : (1) Construct a thin window in the 
counter, usually at one end, or (2) if the /3-ray source can be 
placed in a large chamber where the pressure can be reduced 
to the operating pressure of the counter, the metal cathode of 
the counter can be made of very thin material. As an illus- 
tration of this latter method Smythe and Hemmendinger 10 

10 Smythe, W. R., and Hemmendinger, A., Phys. Rev., SI, 178 (1937). 


have measured the activity of the potassium 40 isotope by 
using a counter with an aluminum wall 0.0254 mm thick and 
letting air at 5.6 cm of mercury pressure into the whole ap- 
paratus after the sample of potassium had been collected. 

The above procedure for /3 rays applies equally well to 
a rays or other heavy charged particles, except that in general 
the windows must be made of thinner material, owing to 
the high energy loss of these particles in passing through 
matter. However, the proportional counter (see page 260) 
or the linear amplifier of Wynn- Williams, 11 as developed by 
Dunning, 12 is usually preferable in detecting these heavy 
particles, since the effects of other ionizing agents are 

For the detection of y rays it is desirable to increase as 
much as possible the number of secondary electrons emitted 
from the walls of the metal tube under the action of the 
radiation. As pointed out by Evans and Mugele, 13 this 
sensitivity can be increased by (1) making the cathode ma- 
terial from one of the heavy elements, such as platinum, 
which increases the absorption of the y rays in the walls 
and produces more secondary electrons, and (2) increasing 
the surface area of the metal forming the cylinder either by 
grooving or by using a fine mesh screen. By these two de- 
vices it is possible to increase the count by a factor of two 
over that given by a plain copper electrode. The usefulness 
of a counter for measuring radiations is illustrated by the 
work of Pohl and Faessler, 14 who have shown that the in- 
tensity of a certain AgK^ radiation required only a few 
minutes to get a measurable response with a G-M counter, 
whereas it was necessary to expose Laue X-ray film to the 
same radiation for 100 hours. 

When light of the proper frequency falls upon the inside of 
the metal tube of a counter, photoelectrons are liberated. 
Such a G-M photoelectric cell becomes a means of detecting 

11 Wynn- Williams, C. E., Roy. Soc., Proc, A, 131, 391 (1931). 

12 Dunning, John R., Rev. Sci. Instruments, 5 387 (1934). 

13 Evans, R. D., and Mugele, R. A., Rev. Sci. Instruments, 7, 441 (1936). 

14 Pohl, M., and Faessler, A., Zeits.f. Physik, 102, 562 (1936). 



[Chap. VII 

very feeble radiations. The metal used for the cylinder in 
the ordinary counter has its photoelectric threshold below 
the region of transmission by the glass and hence does not 
respond to the light falling upon it. For metals of this 
nature the tube may be made to respond to radiations from 
1800 A up to the threshold of the metal by cementing a 
quartz window on one end of the glass tube as shown in 
Fig. 11. The central wire of tungsten should be large enough 

quartz window, ^ tungsten wire >* glass 

ground and cemented on 

Fig. 11. A counter which may be used for photoelectric work. 

to support itself well from one end. The free end should 
terminate in a small ball to eliminate point discharges. 

The whole tube surrounding the metal cylinder could be 
made from quartz, and quartz-glass graded seals used where 
the wires are to be taken through. 

The spectral sensitivity characteristics of counters having 
aluminum, zinc, cadmium, iron, and copper cathodes have 
been determined by Kreuchen 15 in the region 4000 A to 
2540 A. He found that they correspond in all cases for the 
same metals when used in ordinary photoelectric cells. 
Using the three metals zinc, cadmium, and copper in the 
bulk state, Kreuchen 16 also found that the sensitivity was 
increased by activating with hydrogen, but that these metals 
when evaporated showed no increase when treated with 
hydrogen. The photoelectric yield of the activated bulk 
metal and the evaporated metal was the same. 

Very little work has so far been done with cathodes of 
metals that are sensitive in the visible. Locher 17 was one 

15 Kreuchen, K. H., Zeits. f. Physik, 94, 549 (1935). 

16 Kreuchen, K. H., Zeits. f. Physik, 97, 625 (1935). 

17 Locher, G. L., Phys. Rev., 42, 525 (1932). 


of the first to work on this problem, producing counters 
with cylinders of various metals. He tested the sensitivities 
in the ultraviolet and in the visible, both of the pure metals 
and metals coated with dyes and other foreign substances. 
Kolin 18 has succeeded in evaporating sodium, magnesium, 
and calcium onto the metal cylinder and has attained high 
sensitivities in the visible. Christ oph 19 has done the most 
extensive investigation of the problem so far. Using an 
evaporated coating of calcium, he has determined the 
characteristic of the counter for ultraviolet and visible light 
and finds that consistent behavior can be obtained after 
the counter has had a chance to age. 

It appears that photoelectric G-M counters are less re- 
liable than the photoelectric cell at the present time, but, 
owing to the sensitivity attainable, the former may offer 
a fruitful field of research. 

Methods of measuring the number of counts. The string 
electrometer has been applied successfully by Rutherford 
and others 20 in counting particles up to an average of 1000 
per minute by recording the motion of the fiber on a photo- 
graphic film. This method has the advantage of simplicity 
but is limited to relatively low counts because of the response 
of the electrometer. It also has the disadvantage that the 
reading cannot be obtained immediately. The usual 
method today is to take advantage of the amplifying action 
of various kinds of vacuum and gas tubes, which eventually 
operate a mechanical recorder. 

The conventional circuit for such a recorder in its simplest 
form is shown in Fig. 12. The bias on the grid of the first 
tube is such that the plate current is only partially stopped. 
The bias on the second tube is such that it almost completely 
blocks the plate current. When an ionizing particle passes 
through the counter, the wire collects a negative charge, 
which causes the grid of Ti to go negative. The plate poten- 

18 Kolin, A., Rev. Sci. Instruments, 6, 230 (1935). 

19 Christoph, W., Ann. d. Physik, 28, 47 (1935). 

20 Rutherford, Ernest; Chadwick, James; and Ellis, C. D., Radiation from 
Radioactive Substances, page 52. New York: The Macmillan Company, 1930. 



[Chap. VIT 

tial of Ti rises, thus causing the grid on T 2 to go positive. 
T 2 then passes current, and if the quantity of charge is 
sufficient, the mechanical recorder 21 K will be actuated. 

The value that R must have for the counter to function 
properly will depend on the counter. For a very slow 
counter, that is, one which requires a long time for the 
charges to be collected, it may be necessary for Ri to be as 


Type 2.A5 

R 7 = 

0.5 to 5 x 10 9 ohms 
5 x \0 & ohms 
0.5 x lO 6 ohms 
3000 ohms 
10 5 ohms 
2 x lO 3 ohms 
lO 5 ohms 


R 8 = ZOOO ohms 

Ci= 30to50uuf 
C 2 = 10" 5 L(f 

C a = 10" 5 ^f 

C 3 = C 4 = O.l yaf for low voltage 

K= htgh-impedanc& 

V t 

mechanical recorder 
counter threshold + lOOv 

Fig. 12. 

A. conventional circuit for recording the pulses delivered by a 
G-M counter. 

high as 5 X 10 9 ohms. If the counter is a fast one, that is, 
if it is capable of giving a pulse of short duration, then R 1 
may be as low as 10 5 ohms, and in some cases may be dropped 
to 4 X 10 3 ohms. For these short reaction times, the dura- 
tion of the pulse passed on to the second tube is too short 
for a vacuum-type tube, such as the 2A5, to pass sufficient 
quantity of electricity to actuate the mechanical recorder, 
and it is necessary to use either a pulse-lengthening device 
such as the multivibrator circuit to be described later, or 

21 Mechanical recorders of various resistances are manufactured by the 
Central Scientific Company, Chicago, Illinois. 

Chap. VII] 



a gas tube, for example, the 885. A circuit that is self- 
biasing throughout and designed for a fast counter using 
the 885 is shown in Fig. 13. The action of the second half 
of this circuit will be described later. 

Almost any counter can be made to count by using a 
vacuum tube such as the type 57 or 6C6 22 to help the G-M 
counter extinguish itself. By so doing, it is possible to 

+V ^ +250v — +?50v 

C 4 = 10 "V f 

Cs * 0.2 to. 1.0 juf 

K = high-impedance 

mechanical recorder 
V - counter threshold 
plus lOO volts 

R A = R 2 -R 3 = R 6 =R 7 - 10 s ohms 
R 4 - 2000 ohms 
R 5 = + x 10 5 ohms 
R 8 - 10 4 ohms 

R 9 = 7.3 x 10 4 ohms 
Ci • souuf 
Ca - C 3 = O.lyuf 

Fig. 13. A circuit designed for use with a fast counter. The resistance 
in series with the counter is much lower than in the conventional circuit. 
Because of the short pulse, a trigger-type tube is used to actuate the recorder. 

eliminate such a high resistance as is used in the conventional 
circuit as well as greatly to increase the efficiency at high 
counting rates. There are several types of circuits by which 
this may be accomplished. The first 23 is shown in Fig. 14. 
The cylinder of the counter is connected directly to the grid 
of the first tube, while the potential which is applied to the 
wire is also connected to the plate of the tube through a 

22 The grid potential-plate current characteristic of these tubes makes 
them very desirable for this type of work. A negative 4| volts on the control 
grid are sufficient to block a potential of 1500 volts on the plate if 45 volts are 
used on the screen grid. 

23 Neher, H. V., and Harper, W. W., Phys. Rev., 49, 940 (1936). 



[Chap. VII 

resistance R 2 . The coupling to the next tube is made by 
the usual condenser, except that in this case the pulse de- 
livered is large enough to permit a small capacity C of the 
order of 10~~ 5 microfarad to be used instead of 10~ 3 to 10~ 4 
microfarad, as is used for the couplirg condenser in the con- 
ventional circuit. 

The action of the circuit can be explained briefly as follows : 
The vacuum tube is biased close to the point at which very 
little plate current flows. The full potential of Vi is then 
across both the tube and the counter. When an ionizing 


R x = Z to 10x106 ohms 

R 2 = a x IO& ohms 

50 to lOO ///a f t high- 
voltage, low- leak- 
age condenser 
counter threshold 
♦ 100 volts 

C = 


+45 v +V, 

Fig. 14. A circuit for use with a slow or fast counter using low resistances. 
The radio tube helps the G-M counter extinguish itself. 

particle passes through the counter, positive charges col- 
lected by the cylinder cause the grid to go less negative. 
A current flows, causing a drop in potential across R 2 . 
When this drop becomes sufficient, the discharge in the 
counter will be extinguished, and the circuit recovers itself. 
This recovery is very rapid because of the low values of 
capacity and resistance. With such a circuit it has been 
found possible to count random pulses of 10 5 per minute 
with apparently few being missed. 

The pulse delivered to a second tube will be negative, 
which means that the plate current in this tube must flow 
continuously except when a pulse occurs. If it is desired to 
operate a power tube, then, in order to conserve power, it 
is best to use another tube such as a type 27 inserted between 

Chap. VII] 



>K t 


T\ and T 2 to reverse the direction of the pulse. The nega- 
tive pulse delivered by this circuit is, however, just what is 
wanted on the mixing tubes in case two or more counters 
are connected to count coincidences. 

Another method 24 of using a vacuum tube to help the 
G-M counter to recover itself is shown in Fig. 15. It will 
be noticed that no bias is used 

on the grid, so that there is 
normally a drop of only a few 
volts through the tube. Thus 
the cathode, grid resistance, 
screen voltage supply, and so 
forth, are all at a high positive 
potential. This puts a high 
positive potential on the wire 
of the G-M counter. If this 
potential is above the thresh- 
old value, then when an ion- 
izing particle passes through 
the counter, a negative charge 
collects on the grid, blocking 
off the current in the tube. 
This allows the cathode, grid, 
and so forth, to drop rapidly 
toward ground potential. As 
soon as the potential across 
the counter drops below the 
threshold value and the nega- 
tive charges have flowed off 
across R±, however, the grid 





— ■=• +v 

Ri= 5 x lo 6 ohms 

R 2 = lO 6 ohms 

V t = 45 volts 

V = counter threshold + loo V 

C = 50 to lOO ;u/if 

Fig. 15. Second method of using 
a radio tube to help the counter 
extinguish itself. Since there is no 
bias on the grid of the tube, prac- 
tically the full voltage, V, is across R 2 
and hence also across the counter. 

again takes control, and the circuit rapidly recovers itself 
and is ready for another count. 

It will be observed that the pulse taken from the circuit of 
Fig. 15 is negative. If a positive pulse is desired, a resist- 
ance of perhaps 2 X 10 5 ohms may be placed in the plate 
circuit and a positive pulse to the next stage taken off as 

24 Neher, H. V., and Pickering, W. H., Phys. Rev., 53, 316 (1938). 



[Chap. Vli 






R >HR 

R 3 






shown in Fig. 16. If, in addition, it is desired to eliminate 
the screen supply battery, a bleeder may be used as is also 
shown in Fig. 16. When this resistance is placed between 
the high- voltage supply and the cathode, there is a definite 
end to the plateau, owing to the fact that the potential across 

the counter cannot drop be- 
low a certain value, which is 
determined by the ratio 
R4/R2. Consequently, if the 
high voltage is raised to such 
a point that this minimum 
value is above the threshold, 
the counter will not extin- 
guish itself. 

The advantages of this cir- 
cuit over the preceding one 
are as follows: (1) The cylin- 
der of the counter is grounded, 
which means that shielding is 
unnecessary and insulation 
less troublesome. (2) For 
large counters the reaction 
time is less, since only the 
capacity of the wire, which 
is small, plays a part. (3) 
Either a positive or a nega- 
tive pulse may be taken off. 
(4) The potential across the 
radio tube is not so large. 
The disadvantages are as fol- 
lows : (1) An insulated filament supply must be used. (2) The 
high-voltage supply must be able to stand from 0.5 to 1 
milliampere of current continuously. 

This latter circuit has been applied to some large counters 
for cosmic-ray work. Since the number of counts for a 
given solid angle subtended by counters counting coinci- 
dences goes up with the area of the counters, there is a de- 



R 4 =5x lo 6 ohms 
R a = lo 6 ohms 
R 3 =o.a x 10 6 ohms 
R 4 =05x io 6 ohms 
R 5 =o.2 x 10 6 ohms 

C t =C a = 50 to lOOyujuf 
C 3 =0.1 /jf 

V = counter threshold + lOOv 

Fig. 16. Self -biasing arrangement 
for the screen grid of the circuit in 
Fig. 15. A positive pulse may also 
be taken off when the resistance i? 5 
is inserted. 

Chap. VII] 





^16 copper tube 
■for evacuating 
» crimped 
and sealed 
off with 

spring- brass vane 

Fig. 17. Design of some large counters using the circuit shown in Fig. 15c 


cided advantage in increasing the size of the counter tubes. 
In this instance, copper tubes 5 feet long with a x^-inch 
wall were employed. Some 12 inches of space were used at 
one end to mount two of the radio tubes. The arrangement 
is shown in Fig. 17. The Pyrex bowls which seal the ends 
are commercial transmitting station lead-in bowls, with the 
edges ground to fit inside the 6-inch tubing. Commercial 
argon, 99 per cent pure, was used at a pressure of 7 cm of 
mercury. The counting rate of each counter, which is due 
to cosmic rays and radioactivity in the counter and sur- 
roundings, was about 100 per second. The resolving time 
was 2 X 10~ 5 second with the constants shown in Fig. 15. 
This means that with three counters separated in a horizontal 
plane the accidentals were about 5 per hour. 

Because of the short duration of the pulse in the preceding 
circuits, it is sometimes difficult to operate a mechanical 
recorder by using an amplifying tube such as that shown in 
Fig. 12. It is much more satisfactory to use a gas-filled 
trigger-type tube, such as the argon type 884, 885, or a 
mercury-vapor thyratron. 25 The former are quite inex- 
pensive, have a short de-ionization time, and are preferable 
to the latter. One of the most satisfactory methods for 
using this tube has been devised by Pickering. 26 It is shown in 
Fig. 13 and in Fig. 18B, in which the second type of extin- 
guishing circuit described above feeds the recorder circuit. 
The grid of the 885 is self-biased to a little beyond the 
point at which it can keep control. When the positive pulse 
from circuit A causes the gas discharge to take place in the 
885, current flows through the recorder K, causing it to 
record a count. As the current continues to flow, C 2 

25 Each of these tubes has a very low plate resistance when in the conducting 
state. If the grid is sufficiently negative, no plate current will flow, but as 
soon as the grid potential is raised beyond a certain point, a gaseous discharge 
occurs, and the grid loses complete control. If a resistance is in series with the 
plate, the drop in potential inside the tube becomes approximately the ioniza- 
tion potential of the gas, or about 17 volts in the case of the argon tubes. The 
discharge can be stopped by dropping the plate potential below the ionization 
potential of the gas for a few microseconds. 

26 Pickering, W. H., Rev. Sci. Instruments, 9, 180 (1938). 

Chap. VII] 



charges up, and the potential of the cathode approaches 
that of the plate. However, the grid has remained near 
ground potential, so that the effective bias becomes very 
large. When it becomes large enough, and the drop of po- 
tential across the tube becomes sufficiently small, the gaseous 
discharge ceases, and the circuit returns to its normal state. 

C x - SO jJLjjif 

c z = o.z yuf 

C 3 = O.lyUf 

V 4 = counter threshold 

250 volts 

high -impedance 

mechanical recorder 

R,= 5x 10 s ohms 

R 4 = io 6 ohms 

R 3 = R 4 = R 6 = o.5 x io 6 ohms 

R 5 = o.zs x io© ohms V t = counter threshold + lOOv 

R 7 = IO* ohms V, 

R&= current limiting K 

R9= 8 x I0 4 ohms 

Fig. 18. The circuit of Fig. 16 feeds a self-biased recording circuit. 

The value of C 2 can be varied to suit the impedance of the 
recorder K. For the shortest resolving time, (7 2 should be 
made as small as possible. The resistance R% may or may 
not be necessary, depending on the impedance of the re- 
corder. In any case the instantaneous current through the 
885 should not exceed 0.3 ampere. The resolving time of 
the circuit is usually shorter than that of the mechanical 

Another method of producing a pulse of much longer 
duration than the initial pulse and so operating a mechanical 
recorder is with the so-called multivibrator circuit. This 



[Chap. VII 

circuit will give a square wave form in the output, the volt- 
age swing of the plate being nearly the full potential applied. 
The length of the pulse on the output is independent of the 
length of pulse on the input, provided the latter is shorter 
than the natural pulse length delivered by the circuit. 

Medium Power Output 
T x = Type 6C6 
T 2 = Type 41 
Ri=R*=R 5 - 10 s ohms 
R 3 * 0.5 x lO 6 ohms 
R4 S 6 x lO 4 ohms 
R 6 =0 to 1.5x10* ohms 

depending on recorder K. 
R 7 a 10 6 ohms 
Ci = KT 3 to ICTVxfd 
C* = C 3 * O.l jjfd, 400 volt 
C 4 = lO-* tolCTVfd (see text) 
K - high -impedance 

mechanical recorder 
V t *3 volts 
V a = 60 volts 


Large Power Output 
It ■ Type 6A4 
T* ■ Type 6L6 
Ri=R 2 -10 5 ohms 
R 3 = 10* ohms 
R*=10+ ohms 
Rs = 5 x lo + ohms 
R 6 =0 to 1000 ohms 

depending on recorder K 
R 7 =10 6 ohms 
e t = ur 3 tol0~ 4 >ufd 
C* = 

0.5/ufd, 4oovolt 

10"* to 10-4 ^f d ( see text) 


mechanical recorder 

11 volts 

C 5 
C 4 


V 2 =45 volts 

Fig. 19. Multivibrator circuit. 

This natural pulse length is determined chiefly by the feed- 
back capacity C 4j Fig. 19. The quantity of electricity which 
flows during a pulse is quite sufficient to operate a Cenco 
recorder of either the high- or low-impedance type. If the 
6L6 is used in the second stage, as much as 0.3 to 0.4 ampere 


can be delivered for any predetermined time up to say 0.1 
second, provided the impedance of the output circuit ia 
sufficiently low. In case this type of circuit is desired, in- 
stead of using a thyratron such as the type 885, the arrange- 
ment in Fig. 19 may be substituted directly for the two 
tubes in Fig. 13. It has many applications where a non- 
linear, constant pulse size output is desired. By adjusting 
the grid voltage on the second tube the circuit can be made 
nonsensitive to input pulses less than a certain size. For all 
pulses with a voltage swing beyond this limiting value, the 
size of pulse in the output will be constant, provided the 
pulse length in the input is less than the natural pulse length 
of the output. 

The behavior of the circuit may be described briefly as 
follows: Ti is biased so that it acts as a linear amplifier. 
T 2 is biased to just beyond the cutoff. When a negative 
pulse arrives at d, T\ passes a positive pulse onto T 2 , and 
owing to the large condenser C 3 , the grid of T 2 follows closely 
the plate of T x and hence goes positive by an amount de- 
pending on i2 4 and the type of tube T 2 . As the plate of T 2 
drops in voltage, a negative pulse is passed back onto the 
grid of T 1} causing T 2 to become still more conducting. 
This process continues until the plate of the second tube has 
dropped to within a few volts of the potential of its cathode. 
The grid of 7\ is now far below the potential of its cathode, 
and the charge on C 4 must leak off across R 2 . While this is 
occurring, T 2 is still highly conducting. When the grid of 
Ti returns to a point where the first tube again begins to 
conduct current, the plate of T 2 is allowed to rise in potential, 
and this in turn causes T x to be still more conducting. The 
process is just the reverse of the initial stages of building up 
the pulse. The make and break of the current in the plate 
circuit of the second tube is extremely rapid, each appearing 
to consume less than 10 -6 second, provided there is a pure 
resistance load. The time during which the plate current 
in T 2 remains at its constant value can be determined by C 4 . 
As an illustration, when (7 4 = 2 X 10 -4 microfarad in either 


of the circuits shown in Fig. 19, the length of pulse in the 
output is about 5 X 10 -4 second when a pulse the length of 
which is 10 -5 second is fed to the input. 

When the bias of T 2 is lessened, there comes a point when 
the circuit will become unstable and oscillate. Just before 
this point is reached, the sensitivity becomes extremely high. 
The point at which oscillations begin depends on C 4 — the 
larger C 4 , the greater must be the bias voltage on TV Using 
a type 6C6 and a type 41, with C 4 = 0.5 X 10 -4 microfarad 
and the other constants those given in Fig. 19, a 0.007-volt 
pulse input of 10 -5 second duration results in a full voltage 
swing of the plate. This is a voltage amplification of about 
2.5 X 10 4 . For an output pulse of longer duration, for 
example, 10 -2 second, (7 4 must be about 10~ 3 microfarad. 
In this case, for stable operation, T% must be biased such that 
a 3- to 4-volt pulse is needed on the input. 

The circuit is adaptable for use with either a fast or slow 
G-M counter. If used with a fast counter, the negative 
pulse from the wire of the counter can be fed onto the grid 
of Ti directly or through the condenser C x . In case a slow 
counter is used, Getting 27 has pointed out that such a multi- 
vibrator circuit can be made to extinguish the counter in a 
way similar to the action of the circuits in Figs. 14 and 15. 
For such operation the wire of the counter is connected 
directly to the grid of 7\ with R 2 = and Rx = 4 X 10 6 ohms. 
The cylinder of the counter has a negative potential applied 
equal to the threshold voltage plus 100 volts. The value 
of C 4 is adjusted to the reaction time of the individual counter 
as well as the reaction time of the recorder. A value of 
3 X 10 -4 microfarad is an average. The length of the 
plateau, however, is limited to the voltage swing of the out- 
put tube. 

If the counts per unit time become too large, the mechani- 
cal recorder will miss an appreciable number. It is shown 
on page 299 that if any device can respond to only those 
impulses separated by a time interval greater than r, theo 


Getting, I. A., Phys. Rev., S3, 103 (1938). 


the relative number of pulses missed, if they are spaced at 
random in time, is tN, where N is the average number of 
pulses per unit time. To overcome this difficulty, Wynn- 
Williams 28 has devised a scale-of-two circuit which cuts 
down the number of pulses by a factor of just two. It 
consists of two tubes such as the type 885, each tube being 
discharged by every other pulse. If another scale-of-two 
circuit is connected to one of the tubes of the first circuit, 
each tube of the second will respond alternately to half of 
the original puls3s. Thus, one tube of the second circuit 
responds to only one fourth of the original number of pulses. 
This process of adding more scale-of-two circuits may be 
continued indefinitely, with one tube of the final circuit 
counting 2~ n of the original pulses, where n is the number 
of scale-of-two circuits. 

A diagram of a modified 29 set of two of these scale-of-two 
circuits is shown in Fig. 20. The action is as follows: Let 
tubes 3 and 6 have a gaseous discharge. This state of affairs 
can be obtained by first closing Si and then S 2 . The current 
through tubes 3 and 6 will cause a drop in potential across 
R 2 and R±, which will bias tubes 2 and 5 so that they will 
not glow when S 2 is closed. The ratio of the plate potential 
to the grid potential for the grid to keep control is about 
10 to 1 for the type 885 tube. However, to secure consistent 
action it is best to have this ratio somewhat lower. Ratios 
from 5 to 1 to 8 to 1 are recommended. Let a negative pulse 
be delivered to tube 1. A positive pulse will be passed on to 
tubes 2 and 3. Tube 2 will then become conducting, the 
drop in potential from plate to cathode becoming the ioniza- 
tion potential of the argon, or about 17 volts. The dis- 
charge of tube 2 thus causes a sudden drop in potential across 
Rs which is passed on through C 4 to the plate of tube 3. 
But since tube 3 was conducting, R 9 already had a large drop 
in potential across it, and the additional pulse passed on 

28 Wynn- Williams, C. E., Roy. Soc, Proc, 136, 312 (1932). 

29 Shepherd, William G., and Haxby, Robert O., Rev. Set. Instruments, 7, 
425 (1936). 



[Chap. VII 

through C 4 makes the plate of tube 3 go negative with respect 
to the cathode. The discharge is extinguished, and its grid 
then takes control. Tube 2 is now glowing, and tube 3 is in 
the nonconducting state. When a second negative pulse 
arrives at Ci, the above procedure is just the same except 

R t - R 3 =R S = R 7 =Rio= 10 5 ohms C t = o.ao5,uf 

R 3 = K 4 = 600 to lOOO ohms C z = Cs= o.0005/if 

C 4 = 0.02 >uf 

c 5 - c 6 =* c 7 ■* C 3 = 0.001/xf 

C 5 = O.05 ,uf 

R-8-- R 9 = R„ = R U = 5O0O ohms 

R 6 - 10 4 ohms 

R 13 = 7.5 x JO 4 ohms v^ 8 = u.oa a*t 

AM grid resistors = iO 5 ohms C i0 * 0.2 to 1.0 /if 

Tubes 1,4, = Type 56 K = high -impedance 

Tubes 2.3, 5,6,7, = Type 555 mechanical recorder 

The unit between v, a " and b" may be considered a "scale 
of two" unit. Any number of such units may be addecU 

Fig. 20. Modified Wynn-Williams scale-of-four working into the recording 

unit B of Fig. 18. 

that the tubes are reversed. It is obvious, then, that C 5 
will receive a positive pulse only when tube 3 is made non- 
conducting, which will be just half the number of times a 
negative pulse arrives at d. 

The unit from "a" to "b" of Fig. 20 is the same as the unit 
composed of tubes 1, 2, and 3, except for a slight difference 
in circuit constants and in the action of tube 4. This latter 


tube is inverted, so that it passes on a positive pulse as well 
as responds to a positive pulse but is relatively insensitive 
to negative pulses. The number of pulses which appear at 
C 9 is then only one fourth of the number impressed on C\. 
The unit from (a) to (b) may be considered a scale-of-two 
unit, and any number may be added. The final output is 
made to actuate a self-biased recording circuit as shown in 
Fig. 18B. 

Tubes 1 and 4 act as one-way valves, keeping negative 
pulses from passing through. This makes for more con- 
sistent action, since a large negative potential applied to the 
grid of an 885 will sometimes cause it to extinguish. To 
maintain stability, it is necessary to have the time constant 
of the plate circuit greater than that of the grid circuit. 

Still another method is available for counting impulses 
delivered at a high rate. By using a circuit somewhat 
similar to the scale-of-two, Hunt 30 has devised a scheme to 
obtain a pulsating direct current derived from charging and 
discharging condensers by means of type 885 tubes. A 
micro- or milliammeter gives a reading proportional to the 
average rate of arrival of the pulses. By choosing the proper 
capacities and resistances, this direct-frequency meter can 
be used in counting random pulses up to 10 5 per minute. 
The apparatus is conveniently calibrated with a beat- 
frequency oscillator. 

Coincident circuits. In cosmic-ray work especially, it is 
desirable to record only simultaneous discharges of two or 
more counters. Several means have been devised for ac- 
complishing this, but the one now in almost universal use 
is that of Rossi. 31 It may be applied to any number of 
counters. The circuit is shown in Fig. 21. The principle of 
operation is as follows: The plates of all tubes are con- 
nected in parallel across a high resistance E 4 and a source 
of potential. The grid of each tube is normally at the same 
potential as the cathode, so that the drop in potential 

30 Hunt, Frederick V., Rev. Sci. Instruments, 6, 43 (1935). 
» Rossi, B., Nature, 125, 636 (1930). 



[Chap. VII 

across any one in the static condition is small compared with 
the drop across R±. Under these conditions, if a negative 
pulse arrives at d, tube T\ will instantaneously have a much 
higher resistance. However, since T 2 and T$ are in parallel 
with T h there will be very little effect on the current through 
i£ 4 , and hence the drop in potential across R± will be prac- 
tically unaltered. The same holds if two tubes are affected, 
say Ti and T 2) for even though these suddenly assume a high 
resistance due to their grids simultaneously being made more 


I negative I n 

| pulse | 



— 1 — ^2 







- — \ {(positive 



— +80v +Z50v 

R 1 =R 2 =R 3 =2.5xio 5 ohms R-, = lx lo y ohms 

C A = C Z =C3= 50 toloo/i/if C4=0.ooiymf 

Fig. 21. Coincidence circuit of Rossi. Any number of tubes may be used 
in this parallel arrangement. 

negative, the third one still has a low resistance compared 
with R i} and the resultant effect passed on through C 4 will 
still be small. However, if all three grids go more negative 
simultaneously, then the potential at (7 4 rises, and a positive 
pulse is passed to the output. 

With the circuit constants given in Fig. 21 the direct- 
current resistance of each of the 57 or 6C6 tubes is 4000 ohms. 
The maximum possible voltage change to the output when 
one tube only receives a negative pulse is 0.8 volt, and when 
two tubes only are so affected, the maximum possible change 
is 2.8 volts. However, when the grids of all three tubes go 
negative simultaneously, the maximum change can be as 

Chap. VII] 



large as several hundred volts. By suitable adjustment of 
C 4 , or of the grid potential on the first tube in the output, it 
is easy to rule out completely the singles and doubles and 
record only the triples. 

For cosmic-ray work in which doubles, triples, quadruples, 
or any other number of coincidences are to be recorded, the 


B 3 

A 4 =A a =A n = G-M tube connected with vacuum tube. as 
shown in Fig's. 13, 14 or 15 

E> 1 =IB a =B ft = Amplifying" tube such as shown 
in Fig. £1 

C -Suitable mixing tube such as is shown in Tio;.18B 

Fig. 22. Schematic diagram for any number of G-M counters arranged to 
count coincidences. 

following assembly, shown diagrammatically in Fig. 22, can 
be recommended: 

1. Use the required number of high-speed counting cir- 
cuits shown in Figs. 13, 14, or 15. 

2. Connect the outputs through a capacity to the grids of 
the coincidence tubes shown in Fig. 21. 

3. Use the output of this coincident circuit to operate the 
circuit shown in Fig. 18B. 



[Chap. VII 

High-voltage sources. Although batteries furnish an 
ideal source of high potential for counters, the expense in- 
volved usually prohibits their use. High- volt age direct- 
current generators can also be used, but the actual power 
required for the operation of a counter is so small that here 
again the expense is unjustified. The simplest and most 
practical method is to rectify alternating current after the 
potential has been increased with a transformer to the de- 
sired voltage, and then smooth the pulsating output with a 
condenser or, if much current is to be drawn, with one or 
more condensers and chokes. 

In most cases half -wave rectification is sufficient, since the 
actual current drain for a Geiger counter is usually small. 

Half-Wave Rectifier 

Type 866 

~~ output 

Fig. 23. 

C = O.l to 1.0 yuf 

For many purposes where the current drain is small, half-wave 
rectification is sufficient. 

A simple, inexpensive rectifier is shown in Fig. 23, where a 
type 866 mercury-vapor tube 32 allows the condenser, C, to 
charge up to the peak voltage supplied by the transformer. 
It is necessary to have a filament transformer capable of 
withstanding the required potential if the negative side is 
grounded. The condenser C also has across it the full output 
potential. Its capacity need be no larger than 0.1 micro- 
farad if only current to supply the counter is drawn. The 
amount of ripple for a current / can be computed approxi- 
mately from 







32 The type 866 mercury- vapor rectifier is an inexpensive tube which has 
ample current-carrying capacity and is rated for an inverse peak voltage of 
7500. It requires a filament supply of 2.5 volts and 5 amperes. 

Chap. VII] 



Full -Wave Rectifier 

f \T>pc 

( 1866 

L - 30 henry choke 

1 C v >r 1 

L_!r T Cl 


Fig. 24. If power is to be supplied, it is best to use full-wave rectification 
with a suitable filter. 

where AV is the fluctuation in the voltage output, R the 
resistance across the output, and n the number of pulses 
supplied to the condenser, C, per second. 

In case it is desired to draw much current from the output, 
it is best to rectify both halves of the alternating-current 
wave. Two 866 tubes may be used as shown in Fig. 24. 
In this case a transformer with a center-tapped secondary 
winding and one insulated filament transformer are neces- 
sary. The filter consists of two condensers of capacity of 
1 to 2 microfarads and a 30-henry choke. The output of 
such a rectifier and filter unit will have less than a 1 per cent 
ripple for 60-cycle current when the current drain does not 
exceed 10 milliamperes. 

If a rectified voltage is wanted which is greater than the 
peak potential supplied by the transformer, a voltage- 
doubling circuit such as shown in Fig. 25 may be used. The 

Voltage Doubler 




c a 




— o *• 

C x = C z = 0.1 to l.o juf 

Fig. 25. The above arrangement will double the peak voltage supplied by 
the transformer. Two separate filament supplies must be used. 



[Chap. VII 

voltage output will be double the peak voltage available 
from the transformer. The circuit shown employs two 
type 866 mercury-vapor tubes. 

If it is desired to regulate the output voltage, a voltage 
regulator such as will be described later can be used. These 
devices will also take out the ripple, provided the minimum 
at any time does not fall below the stabilized voltage. 

Voltage regulators. Frequently it is desirable to maintain 
a constant voltage, for example, when working with a pro- 
portional Geiger counter. 
Several schemes 33 have been 
devised for accomplishing 
this, but one of the simplest 
is that shown in Fig. 26. The 
action is as follows: As the 
input potential is raised, no 
current flows through the 57, 
owing to the negative bias, 
until the output potential 
reaches a point where the 
grid is at about —3 volts 
with respect to the cathode. 
As the input potential is 
raised still farther, the out- 
put potential at first goes up 
slightly, and then, because of 
the drop in potential across 
Rt, reaches a maximum and 
finally falls. If g m is the mutual conductance of the tube, 
the change of output voltage V with input V t can be ex- 
pressed as 

Ri = 0.2 toZ aIO 6 ohms 
R z " 4 to 20 x 10 6 ohms 
R 3 - Z x 10° ohms 
R 4 = 1 to Z x 10* ohms 
V t = 90 volts 
V 2 = 45 volts 

Fig. 26. Simple type of voltage 
stabilizer which is suitable for po- 
tentials up to 4000 volts. If it is 
desired to stabilize voltages higher 
than this, a pentode designed for 
higher potentials must be used. 

33 See the following: 

Ashworth, J. A., and Muzon, J. C, Rev. Sci. Instruments, 8, 127 (1937). 

Evans, R. D., Rev. Sci. Instruments, 5, 371 (1934). 

Gingrich, N. S., Rev. Sci. Instruments, 7, 207 (1936). 

Richards, L. A., Rev. Sci. Instruments, 4, 479 (1933). 

Street, J. E., and Johnson, T. H., Frank. Inst., J., 214, 155 (1932). 

Webster, H. C, Cambridge Phil. Soc, Proc, 28, 121 (1931-1932). 

Chap. VII] 



dV _ R 2 -\- R$ — RzR\g m 
dV t " Ri + R 2 + Rz + flifla&n" 

If #4 = 0, Ei = 2 X 10 6 ohms, R 2 = 20 X 10 6 ohms, 
# 3 = 2 X 10 6 ohms, Fi = 90 volts, and V 2 = 45 volts, the 
regulation is about 1 per cent; that is, the change of output 
voltage is only 0.01 of the change of the input voltage. 
With R± — 15,000 ohms, and the other quantities the same 
as before, the maximum occurs experimentally at about 
2000 volts input and 1000 volts output, and there is less than 
1 volt change in the output when the input voltage is changed 





K t =Sx io 6 ohms 
R ? = 20xio 6 ohm5 
R a - Z x lo 6 ohms 



1500 2 000 

voltage —input 

Fig. 27. Typical performance curve for the circuit shown in Fig. 26. With 
J?4 = 0, the regulation is about 1 per cent. 

from 1500 to 2500 volts. Experimental results using the 
above circuit constants are shown in Fig. 27. As J? 4 is in- 
creased, the maximum becomes sharper and moves down to 
lower voltages. 

If it is desired to draw current from the output and still 
maintain a constant voltage supply, the above circuit is not 
satisfactory when more than a fraction of a milliampere of 
current is used. It is possible, by using the constant current 
characteristic of another pentode in conjunction with the 
above circuit, to keep a constant voltage output when the 
current is changed from to 1 milliampere. 



[Chap. VII 

The circuit is shown in Fig. 28. The action may be de- 
scribed as follows: As the input voltage is raised, a point 
is reached where T\ becomes conducting,, depending upon the 
ratio of R 2 to R$. Until this time, T 2 has been in a highly 

conducting state, with a di- 
rect-current resistance prob- 
ably less than 1000 ohms, 
because of the +45 volts on 
the grid. As soon as T\ be- 
comes conducting, the drop 
in potential across R± be- 
comes approximately F 3 + 3 
volts and tends to remain at 
this constant value. As the 
input voltage is still further 
raised, T 2 continues to carry 
current but acts as a con- 
stant current device, and the 
voltage is stabilized by the 
action of T h as in the previ- 
ous circuit. If now current 
is drawn from the output, 
the immediate tendency is 
for the grid of Ti to go more 
negative with respect to the 
cathode. This makes T\ less 
conducting, resulting in a 
less negative potential on the 
grid of T 2 . Thus T 2 be- 
comes more conducting to 
supply the current delivered to the output. 

The performance of the circuit is illustrated in Fig. 29, 
where the circuit constants were those in Fig. 28 and 
E 3 = 2 X 10 5 ohms. The mutual interaction of the two 
tubes makes for a much more constant voltage regulation at 
all times than could be had with one tube. Experimentally 
there was less than 0.1 volt change in the output of 1039.5 

Ri ■ Zx lo® ohm5 

R a --2x lO 6 ohms, wire Wound 

R3 = variable, wire wound 

R* * lo 5 ohms 

Vi = 90 volts 

V* =V 3 =V 4 = 45 volts 

Fig. 28. The above combination 
of two type 57 tubes permits excellent 
voltage stabilization for current 
drains up to 1 milliampere. Drifts 
may be as low as 0.1 volt per hour in 
the output. Potentials from several 
hundred to several thousand volts 
can be stabilized with this circuit 
using the type 57 tubes. 

Chap. VII] 



volts when the input changed from 1050 to 2500 volts. A 
change from to 1 milliampere drain at any input voltage 
above the stabilized value changes the output voltage by- 
less than 0.2 volt. When more current than this is drawn, 
the constant voltage characteristic of the circuit gradually 

None of the resistances or potentials in the circuit are 
critical except Vi and the ratio of R 2 to R z . For constancy, 
R 2 and R z should be wire wound and kept at the same 
temperature. If Vi is supplied by new "B" batteries of the 
dry-cell type, very satisfactory results will be obtained, 



1 1 1 1 


y^ ^^* cucren 

+ ■> i i 

+ r + + 
+■ + + + 

rrent =• 1.6 m a 

i- + + + i- 

t = o.8 m a 

t + t- * - 


/ ^ current - 
| + +• + + 

*^« current — O.C 

+ + + ♦ 

: ; : : 

i i ii 

-- o.4- m a 

c + + 

» m a 

+ + + + ■ 

+ -» + + - 

-■- +■ + t- 
+ ■* t t 


1500 ZOOO 

voltage -input 


Fig. 29. Typical performance curves for the circuit shown in Fig. 28. At 
no current drain there is less than 0.1 volt change in the output of 1039.5 volts 
when the input changes from 1050 to 2500 volts. 

since their temperature coefficient is exceedingly small. 
The heater supply of T 2 is not at all critical, and a change of 
50 per cent in the power input changes the output voltage of 
the circuit less than 0.1 volt. However, a change of 50 per 
cent in the power input to the heater of Ti changes the out- 
put voltage by about 5 volts in 1000. Extended tests showed 
that after the first half hour, drifts may amount in the ex- 
treme cases to 1 volt per hour but may be as small as 0.1 volt 
per hour. 

It will be noted that the output voltage is within a few 
volts of the input voltage until the constant voltage region 


begins, and that thereafter the power dissipation is in a radio 
tube and not in resistances. By changing R 3 (Fig. 28) it is 
possible to achieve equal performance of the circuit from 
several hundred to several thousand volts. 

This type of constant voltage device is a valuable aid in 
eliminating the ripple from rectified alternating current. A 
condenser of low capacity can be used in the filter, and these 
circuits will take out the remainder of the ripple, provided the 
lowest potential reached is not below the stabilized voltage. 

Although the type 57 and similar tubes are rated by the 
manufacturer at about 250 volts on the plate, much higher 
voltages than this may be applied if the wattage dissipation 
is kept low. The limiting factor is usually sparking over in 
the base of the tube. Almost all of the tubes of this type 
will stand 2000 volts on the plate, and many of them will 
not break down under 4000 to 5000 volts. 

Discussion of probabilities and errors in Geiger counter 
work. Time between individual particles. If the particles 
are all independent of one another, they arrive at random, 
and the laws of probability can be applied. Assuming a 
constant source of radiation, the probability of finding a 
time interval between t and t + dt is given by 34 


Prft = le J dt, (1) 


where t is the average value of the time interval. Then the 
probability of finding a time interval between t\ and t 2 is 

given by 

- h — — — — 

P t dt = e ~* - e \ ( 2 ) 

In particular, if we want to know the probability of finding 
a time interval equal to or less than the average time inter- 
val t, we get (1 — l/e)= 0.632, and the probability of 
finding a time interval between I and infinity is 0.368. In 

34 See: Handbuch der Exp. Physik, XV, 786 (1928). Wein, W., and Harms, 
F., editors, Leipzig (1928). 


the case of a mechanical recorder which will not respond to 
pulses separated by a time less than r, it is possible to find 
the average number of counts missed as follows: Let the 
average time between pulses, t, be large compared with r. 
Then Eq. 2 gives the probability of finding a time interval 
less than r. To the first order of approximation this 
probability is t/1. Thus, if N is the total number of par- 
ticles counted, the mean error in the count will be Nr/t, and 
the mean relative error will be r/t. 

Number of particles in a given time. If the number of 
particles from a constant source of radiation is counted for 
a certain length of time and compared with the number 
counted again for the same length of time, the two values 
will, in general, be different. The relative error of a single 
set of counts will, of course, decrease as the number of counts 
is increased. 

If n is the average number of particles arriving in a certain 
time, as determined by a long period of counting, and n is 
the actual number arriving in this time, the probability of 
finding this number n is given by Poisson's law 35 

P. = ^f- (3) 

Thus, if by counting a large number of particles, it is 
found that on the average there are 100 per minute from a 
certain source, then the probability that in this same time 
100 will actually be counted is 0.04, and the probability of 
50 being counted is only about 10~ 8 . There is a certain 
probability of any number being counted, but obviously 


Error in a single count of n particles. If the mean or root 
mean square error, e m , is defined by 

2> - n)V> 

2 M 

35 Bateman, H., Phil. Mag., 20, 704 (1910). 


then application to Poisson's law gives 

«. = {n)v\ (4) 

and the probable error 36 is 0.676™ = 0.67 n 1/2 , since for large 
values, n differs from n by only a small amount. The mean 
relative error, therefore, is e m /n, or n~ 1/2 . In order to have a 
probable error of 1 per cent in a single set of counts, it is, 
therefore, necessary to count 4300 particles; and to have a 
probable error of 0.1 per cent, 4.5 X 10 5 particles must be 

Error introduced by background. If a single counter is 
used to measure the activity of a source of radiation which is 
comparable with the natural count of the counter due to 
background, it is important to know the effect of the back- 
ground upon the accuracy of the measurements. 

If the error of one set of counts is ei, and the error in an- 
other set of counts is e 2 , then the error of the sum or difference 

WQ1 be e = ( €l 2 + 62 2 ) 1 / 2 . (5) 

Consequently, if there are Ni counts due to a certain 
radiation plus the background and N 2 counts due to the 
background only, the mean error of the difference which is 
the effect of the source being measured, is (Ni 4- N 2 ) 1/2 , so 
that the relative mean error is (Ni + N 2 ) 1/2 /(Ni — N 2 ), 
and the relative probable error becomes 37 

= 0.67 7 A , ,7, • (6) 

Nt - N 2 (iVi - N 2 ) 

As an example, if the counting when the source to be meas- 
ured is present is twice what it is when only the background 
is being measured, then it is necessary to count 6 X 4500 = 
27,000 counts with the source present to reduce the probable 
error of the difference to 1 per cent. This is six times as 
many counts as would be needed if no background were 
present. The counting time will be three times as long. 

36 See any book on errors for the relation between mean and probable errors. 

37 See also Evans, R. D., and Mugele, R. A. ? Rev. Sci. Instruments, 7, 441 


In addition, half of this number must be counted when only 
the background is present, so that altogether 9 X 4500 = 
40,500 counts must be made, which will take six times as 
long as if there were no background. 

Errors due to accidentals in counting coincidences. If two 
counters are used to count coincidences between them, it 
will be found that even when the counters are separated by 
great distances in a horizontal direction, so that cosmic rays 
do not contribute to the coincidences, there remains a back- 
ground of counts. These " accidentals " must usually be 
taken into account, especially when the real coincidences 
become of the same order of magnitude. 

In a coincidence circuit using two tubes, if a pulse arrives 
at one amplifying tube within a certain time r either before 
or after a similar pulse arrives at the other tube, where r is 
the resolving time, a coincidence will be recorded. There is 
a certain probability that two unrelated pulses will thus be 
recorded as a coincidence, which, of course, is spurious. 

Let the time widths of the pulses from each counter be 
equal. Then the resolving time will be defined as that time 
width of a pulse which will just respond as a coincidence 
when the peak of another similar pulse falls without a time 
2r of the first. 

Let there be an average of Ni pulses per second from one 
counter and an average of 1 per second from another counter. 
Then the probability that one will fall within a time width 
of one of the Ni pulses will be 2tN\) and if there are 2 per 
second on the average from the one counter, then the num- 
ber of accidentals will be 2(2tNi), and so forth; and for N 2 
per second the number of accidentals will be on the average 

An = 2tNiN*. (7) 

For the case of three counters connected to count triple 
coincidences, it is easily shown that the number of acci- 
dentals per second is given by 

A 123 = SrW^Ns, (8) 

when the counters are all separated in such a way that there 


are no real coincidences between any of the counters. The 
generalization of Eq. 8 for any number of counters n con- 
nected to count coincidences is 

An . . .„ = nr n - l N n , (9) 

where it is assumed that N and r are the same for all coun- 
ters. This last equation provides a ready means of deter- 
mining the resolving time of the counter circuit. It is 
necessary only to separate the counters in a horizontal plane 
to such distances that the number of real coincidences be- 
tween any two due to cosmic rays coming in near the horizon 
is small compared with the true accidentals, and to record 
the accidentals as well as the counting rate of one of the n 
similar counters. 

In case the r's and N's are all different, then, as 
Eckart and Shonka 38 have shown, the generalized expression 
becomes : 

A 12 . . .„ = NJf* . . . NnTin . . .r/i + - + . . . -Y 

\Ti T 2 T n J 

Consistency of data. To check whether or not a counter 
set is operating properly, it is usually desirable to compute 
the probable error of the final result in two different ways. 
If these agree in general to within the required limits, then 
it may be assumed that the counters have been working 
consistently and that instrumental fluctuations have been 

Let the mean error in a single determination of Ni counts 
be €i, the mean error in another determination of N2 counts 
be € 2 , and so forth. Then by an extension of Eq. 5 the mean 
error of the result of n determinations is 

e m = ( ei 2 + € 2 2 + . . . e*Y'\ 

or, by Eq. 4, 

^ = (Nx + N 2 + . . . N n y\ 

and the probable error is 

e p = 0.67 (tfi + N 2 + . . .N n y". 

38 Eckart, Carl, and Shonka, Francis R., Phys. Rev., 53, 752 (1938). 

Chap. VII] 



On the other hand, if 2Vi, N 2 , and so forth, are taken over 
equal periods of time, the average value for this period of 

time will be 


Ni + N 2 + . . . N„ 

Let the residuals (N - Ni), (N - N 2 ), . . . (N - N n ) be 
denoted by r. Then the probable error of the result will be 

/ Zr 2 V 
\n(n - I)) 

0.67 4 , 

i(n - 1), 

If there are instrumental fluctuations entering, then e/ is 
usually greater than e P . If e/ is approximately equal to 
e p , it can be safely assumed that the counters are working 
consistently, for there is an even chance that the actual error 
will be greater than that computed, 1 chance in 4.6 that it 
will be greater than twice that computed, and only 1 chance 
in 22 that it will be greater than three times that computed. 


Test on the Consistent Behavior of Two G-M Counters Counting 
Coincidences. N is the Number of Counts per Hour Taken 
with an Automatic Camera, and r the Deviation 
from the Mean or the Residual 



r 2 



r 2 


+ 5 



+ 38 



- 21 



- 36 



+ 75 



- 122 






+ 29 



+ 50 



- 62 






+ 146 






- 14 



- 15 



+ 127 



+ 19 



- 59 



- 4 



+ 41 


2N = 67,083 Sr 2 = 86,021 
iV av . = 3354.1 

Numerical example. The data shown in Table I were 
found with two large G-M counters counting coincidences. 


by taking readings every hour with an automatic camera. 
The probable error computed from the residuals is therefore 

(y r 2 \l/2 
( 7t ) = 10.1, 

n{n - 1)/ 

and the number of counts per hour with the probable error is 

3354.1 ± 10.1. 

Computed from the number of counts, the probable error 
in 67,083 is 0.67(67,083) 1/2 = 174, and the number of counts 
per hour with the probable error can be written 

3354.1 ± 8.7. 

It will be noticed that the probable errors computed in 
these two ways are nearly equal, although that computed 
from the residuals of each hourly reading is somewhat 
larger. However, there is no systematic trend in the data, 
there being nearly equal positive and negative residuals. 
An application of Chauvenet's 39 criterion to these data shows 
that a single residual must be larger than 150 in order to be 
rejected. No residual in the example should, therefore, be 
discarded. Consequently it may be concluded that there 
are no appreciable instrumental fluctuations entering into 
the result. 

39 Palmer, Albert de Forest, Theory of Measurements, page 127. New York: 
McGraw-Hill Book Company, 1912. 


Vacuum Thermopiles and the Measurement 
of Radiant Energy 


C. Hawley Cartwright and John Strong 

\ RADIOMETRIC instrument consists of a blackened 
-^*- receiver, which is heated by the radiant energy to be 
measured. The instrument is provided with some physical 
means for measuring the rise in temperature of the receiver 
produced by the radiant energy. For the most delicate 
measurements the means employed must be responsive to a 
rise of temperature of the order of a few millionths of a 

In comparison with other methods of measuring light 
intensity, a radiometric instrument is characterized by the 
direct and simple way in which the response depends on the 
intensity of the light; the relation between these two quanti- 
ties is linear. Also, the instrument is generally characterized 
by equal sensitivity for all wave lengths. 

For measuring the intensity of radiant energy at wave 
lengths less than 1/jl, radiometric instruments are more 
reliable but less sensitive than other instruments such as 
photoelectric or photographic photometers. Accordingly, a 
radiometric instrument is frequently used as a reference 
instrument for the calibration of photoelectric and photo- 
graphic photometers. In infrared spectroscopy, however, 
the radiometric instrument is the most sensitive instru- 
ment now available. 

When a radiometric instrument is giving its full response 
to a beam of light incident on the receiver, the rate at which 



the heat is lost by the receiver is in equilibrium with the 
rate at which heat is absorbed from the light beam, <£ . 
Inasmuch as the heat lost by the receiver is proportional to 
the produced rise in temperature, AT, we can write 

$ a = LiAT 7 + L 2 AT + L 3 AT + UAT, (1) 

where the L's represent the heat losses in unit time per unit 
temperature change. Thus, L x represents the loss of heat 
by radiation from the receiver, L 2 the loss by air conduction, 
L 3 the loss by conduction through members touching the 
receiver, and L 4 any other means of losing heat, such as, in 
the case of a thermopile, Peltier heat loss. Obviously, it 
is desirable to have the L's small, and for this reason the 
energy is to be concentrated onto a small receiver to reduce 
L x , Furthermore, the receiver is usually mounted in a 
high vacuum in order to make L 2 vanish. 

The response of the instrument is determined by the mag- 
nitude of AT, and different radiometric instruments are 
characterized by the manner in which AT is measured. 

A thermopile measures AT* by means of one or more 
thermoelectric junctions attached to the receiver. 1 

A microradiometer measures AT in the same manner as 
a thermopile. 2 In this instrument, however, the thermo- 

1 Brackett, F. S., and McAlister, E. D., Rev. Sci. Instruments, 1, 191 (1930). 
Burger, H. C, and van Cittert, P. H., Zeits.f. Physik, 66, 210 (1930). 
Coblentz, W. W., Bureau of Standards, Bull., 11, 131 (1914). 
Firestone, F. A., Rev. Sci. Instruments, 1, 630 (1930). 

Johansen, E. S., Ann. d. Physik, 33, 517 (1910); Phys. Zeits., 14, 998 (1913). 

Lebedew, P., Ann. d. Physik, 9, 209 (1902). 

Moll, W. J. H., Inaug. Dissertation Utrecht (1907); Arch. Neerland, 13, 100 

Moll, W. J. H., and Burger, H. C, Zeits.f. Physik, 32, 575 (1925); Phil. 
Mag., 50, 618 to 631 (1925). 

Paschen, F., Ann. d. Physik, 33, 736 (1910). 

Pettit, Edison, and Nicholson, Seth B., Astrophys. J., 56, 327 (1922). 

Pfund, A. H., Phys. Zeits., 13, 870 (1912). 

Rubens, H., Zeits.f. Instrumentenk., 18, 65 (1898). 

2 Boys, C. V., Roy. Soc, Proc, 42, 189 (1887), 44, 96 (1888), 47, 480 (1890); 
Roy. Soc, Phil. Trans., 180 A, 169 (1889). 

Coblentz, W. W., Bureau of Standards, Bull., 2, 479 (1906). 
Paschen, F., Ann. d. Physik, 48, 272 (1893). 


junctions and receiver are attached to the moving system of 
a galvanometer coil, which is suspended on a fine quartz 
fiber. The superiority of the microradiometer over a 
thermopile lies in the fact that, because no outside lead 
wires are required, energy losses in electrical resistance are 
diminished. However, the combination of the thermopile 
and galvanometer makes an instrument which is awkward to 
use in a spectrometer, because it must be protected from 
vibration in its operating position. 

A bolometer consists of a blackened thin metal strip 
with electrical connections. 3 This strip forms the receiver 
for the radiations. It is connected as one arm of a balanced 
Wheatstone bridge. The change in the electrical resistance 
of the strip, as measured by a sensitive bridge galvanometer, 
is a measure of AT. 

A radiometer consists of a system composed of a receiver 
and a mirror which is mounted in a partially evacuated case. 
The system is suspended by a fine quartz fiber. The back of 
the receiver is thermally insulated from the front, so that 
when a beam of light falls on the receiver, the front is heated 
more than the back. 4 

The radiometer is most sensitive at a gas pressure of 
about 0.06 mm of mercury. The gas molecules which strike 
the side of the receiver which is warmed by the radiations 
leave it with a greater velocity than those which strike the 
opposite and cooler side, and therefore a net backward re- 
coil is exerted. This results in a deflection of the system 
until the recoil torque is balanced by the torque arising from 

3 Langley, S. P., Am. Acad., Proc, 16, 342 (1881); Annals of the Astro- 
physical Obs., 4, 45 (1904), 5, 75 (1905). 

Leimbach, G., Ann. d. Physik, 33, 308 (1910). 

4 Abbott, C. G., Astrophys. J., 69, 293 (1929). 

Coblentz, W. W., Bureau of Standards, Bull., 4, 391 (1908), 9, 15 (1913). 

Crookes, Sir William, Roy. Soc, Phil. Trans., 11, 166, 325 (1876). 

Sandyik, O., J.O.S.A., 12, 355 (1926). 

Hettner, G., Zeits.f. Physik, 27, 12 (1924). 

Nichols, E. F., Phys. Rev., 4, 297 (1897). 

Smith, S., Nat. Acad. Sci., Proc, 16, 373 (1930). 

Tear, J. D., Phys. Rev., 23, 641 (1924). 


torsion of the quartz fiber. The deflection of the system, 
as indicated by the mirror, is a measure of the temperature 
difference, AT, between the front and back surfaces of the 

Anyone interested in radiometers will find some of the 
important papers on this subject listed in the footnotes. 
One of the features of the radiometer is its constant sensi- 
tivity. This reproducibility of the deflection is due partly 
to the use of a quartz suspension but mostly to the fact that 
the required pressure (0.06 mm of mercury) is one that is 
easily maintained permanently in a closed-off system. The 
radiometer has been used successfully in the micropho- 
tometer. The application of the radiometer to the micro- 
photometer places but little demand on flexibility. 

When maximum sensitivity is desired for very delicate 
measurements, the problem arises of choosing which type of 
radiometric instrument will be most sensitive, and further, 
which design of a given type will be most sensitive. 

There are conflicting reports on the ultimate sensitivities 
obtainable with the different types of radiometric instru- 
ments. The thermopile is certainly almost as sensitive as 
any other radiometric instrument, and although other in- 
struments might be made slightly more sensitive than 
vacuum thermopiles, they are usually more difficult to con- 
struct and use. 5 Accordingly, in our treatment here, the 
construction details of radiometric instruments other than 
the thermopile will be omitted. Vacuum thermopiles are 
widely used by experimenters in infrared spectroscopy, 
possibly more often than all other types of radiometric 
instruments taken together. 

Construction and evacuation of a sensitive thermopile. 
The construction of a vacuum thermopile of the type shown 
in Fig. 1 will be described here. 6 This thermopile has two 

5 Cartwright, C. H., Physics, 1, 211 (1931). 
Klumb, Hans, Zeits.f. techn. Physik, 17, 279 (1936). 

6 We wish to acknowledge the contributions to this design of Professoi 
Firestone and Mr. Paul Weyrich, of the University of Michigan. 


independent junctions and receivers. Four external leads 
are provided, so that these junctions either can be used 
separately or can be connected together in series or in oppo- 

radiation being" 







1 mm thick 


outside of all 
metal work 

24 copper 
wires ^ 

^16" porcelain tube 
with four holes , 
filed flat on one 
side to permit 
easy evacuation 
of thermocouple 

connections to 
*32 copper wire 

Fig. 1. 



[Chap. VIII 

sition. The thermopile is made compensating by connect- 
ing the junctions in opposition. The receivers are rectangu- 
lar and are placed end to end — an arrangement especially 
suited to spectroscopic investigations. For special prob- 
lems the shape of the receivers as well as other features of the 
design can, of course, be altered. 

A crystalline quartz window is attached with Apiezon 
wax a W." This wax is also used to seal the other joints. 
Apiezon wax "W" is easy to apply and has an extremely low 

vapor pressure — a valuable 
Alundum cement 


feature for maintaining a 
permanent high vacuum. 

A porcelain rod of ^ inch 
in diameter containing four 
holes holds the relatively 
heavy copper wires on 
which the thermo junctions 
are mounted. The project- 
ing copper wires are fast- 
ened together by mica as 
shown in Fig. 1 or by Alun- 
dum cement as shown in 
Fig. 2 so that they will not vibrate. Four flexible and insu- 
lated copper leads are soldered to these heavier copper wires, 
as shown in Fig. 1, and these are brought outside the housing 
through one of the wax seals. 

Fig. 3 shows one method for maintaining a high vacuum 
of better than 10 -4 mm of mercury in a thermopile. The 
Pyrex tube shown here is rilled with activated charcoal. 
The charcoal tube is evacuated and baked for several hours 
to outgas it before the stopcock is closed to isolate the sys- 
tem from the pumps. At first the vacuum will be main- 
tained at better than 10~ 4 mm of mercury for only a few 
hours. However, each time the thermopile is re-evacuated 
the vacuum lasts longer, so that after about five evacuations, 
if the system is tight, the vacuum will remain good for a 
month or so. The vacuum is tested by measuring the sensi- 

Chap. VIIIJ 



tivity of the thermopile under some convenient standard 
condition, such as that of exposing the thermopile to a 
60-watt lamp placed 10 inches away and measuring the 
response of the junction with a relatively insensitive galva- 
nometer. The degree of vacuum obtaining in a thermopile 
should not be tested with a spark, since electrostatic forces 
may destroy the junctions. 




seal with 
Apiezon wax tt W 

Fig. 3. (Use Apiezon wax "W" on the stopcock.) 

Wires for the thermo junctions. One thermoelectric wire 
is made of pure bismuth, and the other is an alloy of bismuth 
and 5 per cent tin. The selection of this combination of 
wires to form thermo junctions has been made after a con- 
sideration of the Wiedemann-Franz coefficients, as well as 
of the thermoelectric powers of various possible combinations, 
including such metals as tellurium and the other bismuth 
alloys. 7 

The resistance of each thermoelectric wire should be at 
least 10 ohms, and the wire should not be longer than 3 mm. 
A bismuth wire 3 mm long with a resistance of 10 ohms has 

7 Cartwright, C. H., Zeits. /. Physik, 92, 153 (1934); Ann. d. Physik, 18, 
656 (1933). 



[Chap. VIII 

a diameter of about 24/*. The bismuth-tin alloy wire should 
have about 20 per cent more electrical resistance than the 
pure bismuth wire, because of the influence of the Wiede- 
mann-Franz coefficient. However, owing to the greater 

specific electrical resistance 
of the bismuth alloy wire, its 
diameter will be about 7/jl 
greater than the diameter of 
the pure bismuth wire. 

Preparation of the alloy 
wires. Thermoelectric wires 
can be purchased from the 
Baker Company, Newark, 
New Jersey, or they may be 
prepared by the Taylor proc- 
ess. To make the wires by 
the Taylor process, the ther- 
moelectric metal is melted 
and sucked up into a thin- 
wall capillary tube of soft glass. (See Fig. 4.) This tube, 
containing the metal as a core, is heated in a small electric 
furnace and drawn out in the manner shown in Fig. 5. The 


Fig. 4. 


capillary filled 
with thermoelectric 
metal * 28 .ga. Nichrome 

ribbon- %" wi d c 

Fig. 5. 


diameter of the wires produced in the composite drawn fibers 
is controlled by the temperature of the furnace and the 
speed of drawing. When the temperature of the furnace is 


properly regulated, the wires obtained are single crystals 
which can be bent and straightened repeatedly without 
breaking. Wires which are brittle should be discarded. 

The glass is removed from the composite fibers with hydro- 
fluoric acid, which dissolves the soft glass readily but scarcely 
corrodes or etches the metal. The hydrofluoric acid, usually 
diluted with a little water to suppress fuming, is conveniently 
held either in a shallow dish which has been coated with 
paraffin or simply in a groove melted in a block of paraffin. 
The wires are withdrawn from the acid with metal forceps 
and washed in a weak solution of Aerosol. 8 (Avoid letting 
the acid come in contact with the fingers.) The wires must 
be freed from all glass or difficulty in cutting and solder- 
ing will be encountered. About 5 minutes in the acid is re- 

The good wires are mounted in flat cigar boxes, one for 
each of the metals. The electrical resistance of each wire 
should be measured and its resistance per unit length noted 
on a small label attached opposite the wire. After an assort- 
ment of wire sizes has been collected and measured, one is 
prepared to proceed with the construction of junctions of 
prescribed characteristics. 

Construction of the junctions. A microscope of about 10- 
power magnification facilitates the manipulation and solder- 
ing of the thermoelectric wires. An erecting binocular 
type giving stereoscopic vision is ideal. 

Fig. 6 illustrates the manner of soldering the thermo- 
electric wires to the copper supporting wires with a hot 
tinned sewing needle. The hot-wire device used for heating 
the needle is electrically heated, the heat being regulated by 
a resistance. The temperature of the tip of the sewing needle 

8 Aerosol or the detergent Dreft, the latter of which is sold in grocery stores, 
has many uses around the laboratory. Besides its usefulness in washing glass, 
aluminum mirrors, and so forth, it can be added to water to decrease its surface 
tension and increase wetting power. This is advisable for washing thermo- 
couple wires, as the solution wets the wires and dissolves the hydrofluoric acid. 
Also, for coating the receivers, the solution with added Dreft has less "attrac- 
tion" due to surface tension, and accordingly there is less danger of destroying 
the work when the brush with its blackening material is applied. 



[Chap. VIII 

can be further controlled by varying the point of contact 
between the hot wire and the tip of the needle. 

Wood's metal is used for 
soldering. A solution of pure 
zinc chloride in distilled 
water is used as flux. After 
the soldering is completed, 
the excess zinc chloride 
should be carefully removed 
with a small brush wet with 
distilled water. 

When the thermoelectric 
wires, which are selected for 
size so that each will be 
about 3 mm long, are soldered 

wires in place 

•tinned areas 
wet with flux 

(ZnCI* + 

distilled H 2 0) 

Fig. 6. 

to the tinned copper supports, they are then "cut" to the 
proper length by touching them with the hot tinned needle 
as shown in Fig. 7. This not only "cuts" the wires but 
tins their ends at the same time. Difficulty with this 
operation will be encountered unless all of the glass has been 
dissolved off the wires. 

These thermoelectric wires 
are now manipulated with a 
cold needle so that their ends 
are in contact. A little flux is 
added to their junction, and 
the soldering is effected by 
heat radiated from the hot- 
wire device. (See Fig. 8.) 
The junction is to be care- 
fully watched. The instant 
to withdraw the heat is indi- 
cated by a slight jerk of the 
tips of the wires due to surface tension of the fused metal. If 
the resistance is too great, each wire is shortened by heating 
the Wood's metal at the base of the wire. Molten Wood's 
metal pulls in the thermocouple wire by surface tension. 

-thermocouple wires 
about ZOjn diameter 

cutting thermoelectric wires 
Fig. 7. 

Chap. VIII] 



The needle is used to heat the Wood's metal. In this way, 
it is easy to construct two junctions with only a fraction of 
an ohm difference in their electrical resistances; and, if the 

quartz fibers are 
attached to the 
Support and the 
wires with 
thin lacquer 

flux 15 added 
to junction 

soldering the junctions 

Fig. 8. 

quartz fibers 

supporting the junctions 
Fig. 9. 

wires used have been taken from the same stock piece of 
bismuth or alloy wire, the sensitivities of the junction will 
match closely. 

Ruggedness in the final thermopile is obtained by the use 
of fine quartz fibers to support the thermoelectric wires and 
attached receivers. The quartz fibers are fastened to the 
copper supporting wires by thin lacquer as illustrated in 
Fig. 9. 

The receivers are made of thin gold foil of about 0.5/x 
thickness. This is considerably thicker than sign painters' 

*Vb lotting 
-^ paper s 

stiffening receivers by 
method of cutting receivers curving them 

These operations are best performed under a 10" 
power microscope. 

Fig. 10. 



[Chap. VIII 

gold leaf. 9 The receivers are cut to size (3 mm by 0.3 mm 
is a convenient size for spectroscopy) on the stage of the 
10X microscope by means of a razor blade as shown in 
Fig. 10. The receivers are strengthened mechanically by 
giving them a cylindrical curvature in the following manner : 
The receiver is placed on a sheet of thin fine-grade paper 
mounted on blotting paper, and a rod of about 0.5 mm in 
diameter is pressed against it. (See Fig. 10.) Gold is 
particularly suitable for receivers because it is easily 

A tiny bit of Wood's metal fused to the junction by radia- 
tion and wetted with flux facilitates attaching the receiver. 

the receivers are 
coated with lampblacK 
mixed with water 
that contains a. 
trace of glue 

soldering receivers *m place blackening the receivers 
Fig. 11. Fig. 12. 

The gold receiver is laid in contact with the thermo junction 
and soldered by heating it with radiation from the hot-wire 
device. (See Fig. 11.) A slight jerk of the receiver indi- 
cates when the heater should be withdrawn. 

After the receivers are soldered in place, they are blackened 
with lampblack or other blackening material with the aid of 
a very small amount of glue as a binder. This mixture is 
applied to the receiver with a small camel's-hair brush as 
shown in Fig. 12. 

9 Gold leaf of the required thickness is prepared by evaporating a proper 
amount of gold in vacuum (see Chapter IV) from a tungsten coil onto a glass 
plate. The film is then washed off the glass with a stream of water. 

Chap. VIII] 



Finally, two quartz fibers are fastened over each receiver 
for added ruggedness. The fibers are so fine and at the same 
time such poor heat conductors that the ruggedness gained 
by their use more than compensates for the negligible heat 
leakage which they introduce. Fig. 13 illustrates the method 
of securing the receivers and shows the completed thermopile. 

quartz fibers are 
\&id over the 
receivers and 
fastened to the 
supports with 
thin lacquer 

securing the receivers 

of completed 

Fig. 13. 

Alternative methods of constructing thermopiles. Some 
experimenters prefer to make the housing for a thermopile 
from blown glass. Fig. 14 shows a popular type of glass 
housing. Fig. 15 shows how the junctions are manipulated 
in the field of the binocular microscope. 

Materials for Thermopile Windows 

Window Material 

Spectral Region for Investigation 

Crystalline quartz 

Ultraviolet to 3.5/* 


Ultraviolet to 9/* 


Ultraviolet to 17/* 


Ultraviolet to 21/* 


Ultraviolet to 30/* 


Ultraviolet to 35/* 

High-melting-point paraffin 

20/* to oo 

Crystalline quartz 

45/* to oo 



ready for use 





to vacuum pumps 

Fig. 14. 

Chap. VIII] 



metal block 
to holdworK 

hand rest 

lis surface 
should be level 
'with the thermo- 

Fig. 15. 

Fig. 16 illustrates a method 

The selection of the proper window material for ths 
thermopile is governed by the spectral region in which it is 
to be used. The appropriate 
choice can be made from the 
data given in Table I. 

High-melting-point paraffin 
for use in the far infrared, 
listed in Table I, should not be 
confused with ordinary low- 
melting-point paraffin. High- 
melting-point paraffin is a 
crystalline material that does 
not deform when it is subjected 
to small stresses. In order to 
obtain strength and at the 
same time have the paraffin win- 
dow very thin, it is advisable 
fco make the Window cylindrical 
of using a tube of paraffin turned out in the lathe. It is 
sufficient to have the cylindrical paraffin window only 1 mm 
thick. Inasmuch as the thermopile cannot be seen through 

the paraffin window, it is neces- 
sary to adjust the receivers to 
the focal point of the radia- 
^ Ss *Jf 'If ifl| tions with the help of the gal- 


Although the Taylor process 
for preparing thermoelectric 
wires is recommended, it is pos- 
sible to obtain wires by the 
process used by Professor A. H. 
Pfund, whereby the molten 
metal is spashed on a plate of 
glass. One may either select 
small wires, that are accident- 
ally formed, or cut wires with a razor blade from the thin 
foil uia.t is also formed. Wires obtained by this method have 

cap turned from 


seal with 

Fig. 16. 



[Chap. VIII 

the disadvantage that, owing to fluctuations in their size, it 
is difficult to make matched junctions with them. 

An alternate method of joining the thermocouple wires 
and attaching the receivers involves welding the thermo- 
electric wires together by means of a condenser discharge. 
The details of this procedure are given in the paper cited 
below. 10 

The receiver may be waxed to the welded thermo junction 
with Apiezon wax "W." This method of attaching the re- 
ceiver yields almost the same sensitivity as soldering. 

It is easier to construct a multiple- junction thermopile 
if one large receiver is waxed to the junctions than to under- 
take the delicate task of soldering separate small receivers 

to each junction. The elec- 

stee-l yoke 

old knife 


trical insulation between the 
junctions of a multiple-junc- 
tion thermopile can be ef- 
fected by coating each junc- 
tion with lacquer before 
applying the wax used for 
holding the receivers. 

Some experimenters con- 
struct thermocouples in an 
order almost opposite to 
that described. The junc- 
tion is formed, the receiver 
is fastened to the junction, 
and, finally, the thermo- 
couple is soldered to the 
supporting wires. 11 This 
procedure is especially suited to the construction of a ther- 
mocouple with a small circular receiver, such as may be 
required for stellar radiometry. For a stellar thermocouple 
the junction may be soldered with a larger bit of Wood's 
metal so that there is formed at the junction a small sphere 

pol ished 


piece of Cal rod 

The thermoelectric wires to 
be welded are placed on the 
marked spot and the blunt 
point is pressed dowry 
on them. 

Fig. 17. 

10 Cartwright, C. H., Rev. Sci. Instruments, 3, 73 (1932). 

11 Firestone, F. A., Rev. Sci. Instruments, 1, 630 (1930). 


of metal, which is then compressed to form a flat receiver of 
circular shape and of the desired diameter. 

Professor Pfund constructs thermocouples by compressing 
the thermoelectric wires together on a plate of polished steel 
that is heated to about 100°C. 12 The receiver can be joined 
to the junction in the same manner. A special device made 
from a knife switch is used for the manipulation as shown in 
Fig. 17. 

For most applications lampblack is suitable for coating the 
receiver, but in special cases it may be preferable to use a 
selective absorbing material for "blackening" the receiver. 13 
Thus, a thermopile used for investigation in the far infra- 
red spectrum between 52^t and 152/jl might have receivers 
"blackened" with powdered glass. For work in the visible 
and ultraviolet spectrum an electrolytic deposit of platinum 
black is particularly suitable. 

The loss of heat by the radiation from a receiver is deter- 
mined primarily by the emission of the receiver in the spec- 
tral region around 10/a (the region in which the maximum 
emission from a black body at room temperature occurs). 
The emissive power of platinum black in the region around 
10/uL is weak (about 20 per cent of that of a black body). 
Thus, the use of platinum black has the effect of reducing the 
heat loss L h so that the receiver is effectively only one fifth 
as great as if the receiver were coated with a material that is 
black for the heat spectrum as well as for the visible spec- 
trum. Besides increasing the sensitivity, this has the further 
advantage of reducing the theoretical number of junctions 
required for the best design. Unblackened silver is sug- 
gested for receivers to be used in the ultraviolet region. 

Fig. 3 illustrates the use of active charcoal for maintaining 
a high vacuum in the thermopile. An alternate method 

12 Pfund, A. H., "Radiation Thermopiles," Rev. Sci. Instruments, 8, 417 

13 Woltersdorff, W., Zeits.f. Physik, 91, 230 (1934). 

Forsythe, W. E., Measurement of Radiant Energy, page 210. New York: 
McGraw-Hill Book Company, 1937. 

Pfund, A. H., J.O.S.A., 23, 375 (1933), 23, 270 (1933). 
Strong, J., Rev, Sci. Instruments, 3, 65 (1932). 


involves the use of calcium as a getter. This method has 
been used by Dr. Pettit of the Mount Wilson Observatory 
and is quite satisfactory. Its use amounts to replacing the 
active charcoal in the thermopile in Fig. 3 with fresh calcium 
filings. These calcium filings are baked out while the tube 
is connected to the pump. Later, from time to time when 
the sensitivity of the thermopile falls off, owing to a decay 
of the vacuum, maximum sensitivity can be re-established 
simply by reheating the calcium. 

The use of sensitive thermopiles. As ordinarily used, the 
radiant energy focused on the active receiving surface of the 
thermopile is interrupted periodically to isolate the effect 
of this radiation from the effect of other radiations falling on 
the receiver. The excursion of the galvanometer resulting 
from interrupting the measured beam is ascribed to changes 
in the temperature of the junctions produced by the radiant 
energy. Considering that delicate measurements may pro- 
duce a change in temperature of only 10 _6 °C., it is necessary 
to interrupt the light rather accurately to compensate for 
the first-order drifts which arise owing to a constant warm- 
ing or cooling of the surroundings of the entire thermopile. 
As a result, just as much time is required for controlling the 
zero position of the galvanometer as for determining the 
deflection produced by the energy being measured. 

It is evident that care is required in selecting the best 
position for the shutter in an optical system. For example, 
it is required that the change in the radiant energy falling on 
the thermopile due to closing the shutter should be the same 
as the change produced by removing the source of the radia- 
tions without changing the position of any object "seen" 
by the thermopile. Otherwise, the variation of radiation 
from closing the shutter may falsify the measurement. The 
shutter is to be put before the entrance slit of the spec- 
trometer rather than after the exit slit in order to mini- 
mize this possibility. 

Compensated thermopiles. While first-order drifts in 
the galvanometer can be eliminated even for an uncom- 


pensated thermopile by properly timing the exposures of the 
thermopile to the radiant energy, second-order drifts (due 
to a change in rate of the drift) can be eliminated only by 
the use of a compensated thermopile. In practice, it is 
difficult to construct a compensating receiver that will effect 
the elimination of more than 90 per cent of the galvanometer 
drift, but further compensation can be achieved by shunting 
an electrical resistance across the most sensitive of the 
junctions, either the active or the compensating ones. The 
junctions to be shunted and the value of the shunt resistance 
are determined experimentally. When the shunt resistance 
has the proper value, severe temperature changes of the 
surroundings of the thermopile housing produce a minimum 
deflection of the galvanometer. If care has been taken in 
constructing a compensated thermopile, the shunting re- 
sistance will be great enough so that the sensitivity of the 
thermopile is not appreciably impaired. One method of 
testing the compensation is to hold a hot soldering iron a few 
centimeters in front of the thermopile. When, for example, 
a particular thermopile of the type shown in Fig. 1 was 
compensated, the galvanometer drift was diminished to a 
twentieth part of the original drift, and it was reduced 
further a hundredfold by the shunting resistance. 

Ordinarily, the energy to be measured is concentrated on 
one receiver; the compensating receiver then acts as an ex- 
ternal resistance in the galvanometer circuit, and therefore 
the deflections are somewhat diminished. In most cases the 
reduction of first- and second-order drifts justifies compen- 
sation and the attendant smaller deflections. 

By another procedure the image of the exit slit of the 
spectrometer covers both receivers, while a shutter in front 
of the entrance slit of the spectrometer obscures first the 
aperture of the half of the slit focused on one receiver and 
then the half focused on the other receiver. 14 Thus the 
area of each of the two receivers is half the area of the slit. 
Theoretically, this scheme is expected to yield about 40 

14 Badger, R. M., J.O.S.A., 15, 370 (1927). 


per cent more sensitivity than the ordinary compensated 
thermopile which has the area of the active receiver, as well 
as that of the compensating one, each equal to the area of the 
slit. In order to realize this 40 per cent gain in another 
but less desirable way, the mirror used for concentrating the 
radiant energy may be tilted periodically, so that the image 
of the exit slit of the spectrometer covers first one receiver 
and then the other. 

Auxiliary apparatus. Ordinarily a galvanometer having a 
period of about 7 seconds and a low resistance of about 10 to 
15 ohms is used with a thermopile. For making delicate 
measurements, the wires leading from the thermopile to the 
galvanometer should be shielded, so that alternating cur- 
rents will not be induced in them by stray electromagnetic 
fields. When the wires are not properly shielded, induced 
alternating currents are, in a sense, rectified by the thermo- 
pile, especially by an uncompensated thermopile, and give a 
spurious galvanometer deflection. 

A simple method of measuring the galvanometer response 
is to observe a well-illuminated scale with a telescope. The 
galvanometer should be arranged so that the scale is at a 
distance of about 5 m. A telescope of about 32-power 
magnification, placed as close as possible to the galva- 
nometer, should be used. With a galvanometer mirror 10 mm 
in diameter, it should be possible to see the millimeter 
divisions so clearly on a scale at a distance of 5 m that de- 
flections on the scale can be estimated to a small fraction 
of a millimeter. 

A lack of definition is often erroneously attributed to the 
galvanometer mirror, but it is usually due to the use of 
optically imperfect glass for the galvanometer window. 
However, there is a limit to the definition attainable, because 
of the finite size of the galvanometer mirror and the effect of 
diffraction. A simple rule is that the scale distance as 
measured in meters must not be greater than the diameter of 
the galvanometer mirror as measured in millimeters. Thus, 
for a scale distance of 5 m, the galvanometer should be at 

Chap. VIII] 



least 5 mm in diameter. About -^ mm deflection at a dis- 
tance of 5 m corresponds to the unavoidable natural fluctua- 
tions in the position of the galvanometer due to Brownian 

The accuracy with which the position of a cross hair on a 
millimeter scale can be estimated is much greater than 




plane-parallel glass plates 
7 mm thick mounted on 
~ vertical pivots 

5 ° crn 


slit about ' auto 

I ljm wide f lamp 


l^O-power reading 
microscope, 8-power 
o bj ect i ve 

A — lever for adjusting parallel plate to compensate for 

the galvanometer deflection 

B — tangent scale —The position of lever A read on 

this scaJe is proportional to the galvanometer deflection. 

C - lever for adjusting "zero" 

O ~ reticule in microscope —suggested forrrT 


image of slit 
in reading position 

Fig. 18. 

might at first be supposed. A standard laboratory experi- 
ment for students at the University of Berlin is to estimate 
the positions of extra marks made on a millimeter scale. 
All of the extra marks are made on a ruling engine, so that 
their positions are accurately known. Although the lines 
are all about T V mm wide, the student is asked to estimate 
the position of each extra line to yi'o mm. In estimating 
these positions, a student seldom makes an error of -^o mm, 


and an experienced observer will have a. probable error for a 
single reading of about 0.03 mm. Accordingly, it is signifi- 
cant to estimate galvanometer readings to ^- mm. 

Fig. 18 shows an ingenious and accurate arrangement 
used by Professor Czerny for determining the magnitude of 
small galvanometer deflections. 15 

Relays. A convenient method of reading galvanometer 
deflections is to use an optical amplifier. Also, when it is 
desirable to record radiometric measurements photographi- 
cally, the primary deflections should be amplified by means 
of some type of relay, and the deflections of a secondary 
galvanometer recorded on moving photographic paper. 

The Moll and Burger thermo-relay may be used for am- 
plifying galvanometer deflections until Brownian motion 
becomes conspicuous. 16 Other amplifiers include the barrier- 
layer photocell amplifier described by Barnes and Matossi 17 
and the thermopile with two triangular-shaped receivers 
described by Cartwright. 18 

The Barnes and Matossi type of relay is made by dividing 
the active surface of a barrier-layer photocell by scratching 
along a diameter so as to make two contiguous semicircular 
areas of active surface. The arrangement of this amplifier 
is illustrated in Fig. 17, Chapter X. Leeds and Northrup 
produce an amplifying galvanometer and photocell combina- 
tion of this type. 19 

The above methods of amplifying galvanometer deflec- 
tions also magnify the drift of the primary galvanometer. 
This is undesirable. Pfund and Hardy have devised a 
resonance radiometer, which tends to "ignore" drift and 
separate it from the response to the measured radiation. 20 

15 Czerny, M., Zeits.f. Physik, 90, 468 (1934). 

Czerny, M., Heins, H., and Woltersdorff, W., Zeits. f. Physik, 95, 262 (1935). 

16 Moll, W. J. H., Phil. Mag., 50, 624 (1925) . The Moll and Burger thermo- 
relay is sold by Kipp and Sonen, Delft, Holland. 

17 Barnes, R. B., and Matossi, R., Zeits.f. Physik, 76, 24 (1932). 

18 Cartwright, C. H., Rev. Sci. Instruments, 3, 221 (1932). 

19 Leeds and Northrup Company, Philadelphia, Pennsylvania. 

20 Hardy, J. D., Rev. Sci. Instruments, 1, 429 (1929), 5, 120 (1934). 
Pfund, A. H., Science, 2, 69 (1929). 


Their scheme is somewhat elaborate and requires the use of 
a tuned pendulum shutter, in addition to two identical gal- 
vanometers. However, the instrument has advantages, es- 
pecially when the thermopile is not adequately protected 
from extraneous thermal effects. Pfund describes the 
resonance radiometer briefly as follows: 

If primary and secondary galvanometers are underdamped and 
adjusted to the same period, then, by interrupting the radiation 
falling on the thermopile with a periodicity corresponding to that 
of the galvanometers, a condition of resonance is set up. As a 
class, resonating systems are characterized by high sensitivity 
for " tuned' ' periodic disturbances and by indifference to random 

This indifference to random disturbances unfortunately 
does not include Brownian motions of the primary galva- 
nometer. Hardy has measured the effect of the Brownian 
motion on the resonance radiometer and has found that the 
fluctuations in the deflection of the secondary galvanometer 
are magnified in accord with theoretical predictions for fluctu- 
ations due to Brownian motion. 21 Nevertheless, Hardy feels 
that delicate measurement to the limit set by these effects is 
definitely facilitated by the use of the resonance radiometer. 
The slowness of the resonance radiometer (it takes about 90 
seconds to make a measurement) is one of its disadvantages. 

Firestone 22 has made an ingenious variation from the 
Pfund scheme. It depends on charging and discharging a 
condenser through the secondary galvanometer with a 
circuit controlled by the amplified thermocouple current. 
A photocell amplifier is used. Naturally, as the output 
galvanometer circuit has infinite ohmic resistance, owing to 
the condenser in the circuit, no net current can flow, and 
consequently all deflections are excursions about an unchang- 
ing zero position. 

We have emphasized the importance of using a com- 
pensated thermopile to diminish galvanometer drifts as well 

21 &ie also Van Lear. G. A.. Jr.. Rev. Set. Instruments, 4, 21 (1933). 

22 Firestone, P. A., Rev. Sci. Instruments, 3, 163 (1932). 



[Chap. VIII 

as to make the circuit electrically insensitive to high-fre- 
quency electromagnetic radiations. For the most delicate 
measurements, it is also necessary to have the galvanometer 

pie pans filled 
with oil 

'/a" plywood 

wedges may be 
plaeed here 
while making^ 
adjustments N^ 


when in use the 
rods must not 
touch the sides 
of these holes 

&" plywood 

leveling nuts 

Fig. 19. Vibrationless support for a galvanometer. The plywood triangle, 
on which the galvanometer stands, should be loaded with lead weights until 
the natural oscillations have a period of about 2 seconds. 

free from mechanical vibrations. This can be accomplished 
by the use of a vibrationless support such as the type shown 
in Fig. 19. The description of this vibrationless support is 
given in Chapter XIV. 


Construction of thermojunctions by evaporation and 
sputtering. There are other applications of thermopiles and 
thermocouples, such as their use for vacuum manometers, 
for measuring alternating currents, for measuring sound 
intensities, for magnifying deflections of a spot of light in 
thermo-relays, and for total-radiation pyrometers. We 
cannot go into all these applications in detail, but the 
present chapter and the references cited should serve to 
guide an experimenter in these fields. The construction 
of thermopiles by evaporation and sputtering, however, 
warrants a description. 

Thermopiles made from films of the thermoelectrically 
active metals, produced by evaporation or sputtering, can 
be constructed having a very low heat capacity, so low, in 
fact, that they will respond to the adiabatic heating produced 
by separate sound waves of 5000 cycles frequency. 23 

One of the metal films used is bismuth and the other is 
antimony. The foundation on which the metal films are 
deposited must be extremely thin and strong. For this 
purpose, glass, mica, or lacquer films are used. 

When a soft-glass tube is fused at one end and strongly 
blown out with air pressure so as to expand and explode a 
thin bulb, the shattered bulb wall yields fine ribbons of glass 
about 1 or 2 mm wide and 1 or 2 cm long. These ribbons 
are of such a thickness as to give interference colors and make 
a suitable foundation for evaporated thermocouples. 

When a mica sheet is rolled upon a stick of about 2 mm 
in diameter so that one of the principal directions is parallel 
to the stick, it is subject to shearing forces. These forces 
produce cleavages, so that when the sheet is subsequently 
split, bands from 1 to 0.1 mm wide are obtained which, 
judging from their interference colors, are as thin as or 
thinner than l/*. 24 

Films for use as a thermopile base, or for many other 

23 Harris, L., and Johnson, E. A., Rev. Sci. Instruments, 5, 153 (1934). 

24 This is the technique described in Burger, H. C, and van Cittert, P. H., 
Zeits. f. Physik, 66, 210 (1930). 



[Chap. VIII 

purposes, may be made by dropping a thinned solution of 
lacquer onto the surface of a bowl of dust-free distilled 
water. 25 Surface tension causes the drop to spread out, 
forming a liquid film on the water over about half the area 
of the water surface. The lacquer soon becomes solid as the 
solvent evaporates. Fig. 20 shows how these films are 
taken off the water on a metal frame. They are allowed to 
dry after the peripheral area of the film is pulled back any- 
where that it is in contact with the main stretched area. 
The thickness of film desired is controlled by varying the 
dilution of the lacquer before it is dropped on the water. 

wire frame 

me ^? 

drop of 

wire frame double 

(raised) lacquer f ilm 

Fig. 20. 

Extremely thin uniform films are formed on water cooled to 
0°C. Films as thin as 5 X 10~ 6 cm are obtainable. Double 
films formed on a frame as illustrated in Fig. 20 are stronger 
than single films of double thickness, owing to the fact that, 
in the case of double films, weak areas in one film are seldom 
opposite weak areas in the second film. 

When the thermoelectric metal is deposited on the founda- 
tion film by evaporation, the heat of condensation of the 
metal vapor, as well as the heat radiated by the filament 
and absorbed by the film, tends to elevate the temperature 
of the foundation. It is necessary to prevent the tempera- 
ture of the film from rising to a point at which it might be 

26 Harris, L., and Johnson, E. A., Rev. Sci. Instruments, 4, 454 (1933). They 
use methyl and ethyl acetate solvent for 2 parts cellulose acetate and 1 part 
glyptal lacquer at 0° C. to get the strongest films. 

Czerny, M., and Mollet, P., Zeits.f. Physik, 108, 85 (1937). 


destructive: The films are mounted in the evaporation 
chamber in contact with mercury or, better yet, in contact 
with a copper cooling block. 

Following the procedure described by Burger and van 
Cittert 26 bismuth and antimony are used for the thermo- 
j unctions, the bismuth being evaporated to form a strip 
about 1/ul thick, while the antimony is evaporated to form a 
strip of half this thickness. The proper weight of metal to 
be evaporated is determined by a simple calculation using 
Eq. 2 in Chapter IV. The area coated with the metal is 
defined by templates. The bismuth strip, which is evap- 
orated first, is deposited a little beyond the point which is 
to be the center of the junction, say 0.2 mm or so. Then, 
the evaporated antimony strip is allowed to overlap the 
center by an equal amount. The area where the strips over- 
lap forms the junction. The junction is then coated by 
evaporation with bismuth black, antimony black, or zinc 
black over a prescribed area, which is defined by baffles. 

To form an area to which electrical contact may be es- 
tablished, gold is sputtered or evaporated at appropriate 
points on the metal films. The connector wires may then 
be soldered to the gold. 

The bismuth crystals formed in the strip by condensation 
of vapors have their axes perpendicular to the base. This 
crystal orientation results in a thermoelectromotive force 
against antimony of 75 microvolts/ °C. The optimum crys- 
tal orientation, so far unattainable by evaporation, gives a 
thermoelectromotive force of about twice this value. 

Evaporated thermo junctions are especially useful for 
making the Moll and Burger type thermo-relay. Burger 
and van Cittert were able to obtain a sensitivity about two 
and one-half times as great as that obtained with the 
ordinary rolled Moll and Burger element. 

Considerations in thermopile design. The thermopile 
shown in Fig. 1 and described above can be adapted to meet 
most of the needs of an experimenter interested in making 
26 Burger, H. C, and van Cittert, P. H., Zeits. f. Physik, 66, 210 (1930). 


radiometric measurements. Some experimenters, espe- 
cially those intending to make extremely delicate measure- 
ments, will be interested in the theory for the design of 
thermopiles. For example, the experimenter designing a 
vacuum thermopile of a given area has several decisions to 
make. He must decide which metals to select for the 
thermocouple wires and determine whether to make few or 
many junctions. Also, he must decide on the material to 
be used for coating the receivers. Or, he may wish to de- 
sign a thermopile to operate at atmospheric pressure. 

The equations expressing the theoretical dependence of the 
galvanometer response on the number of junctions, area of 
receiver, characteristics of thermoelectric wires, and so forth, 
have been completely developed. 27 Calculations based on 
this theory require a knowledge of the characteristics of the 
thermoelectric wires, namely, their thermoelectric power, 
electrical conductivity, and heat conductivity. The calcu- 
lations also require a knowledge of the optical properties of 
receiving surfaces, such as their emissivity and reflectivity 
for various wave lengths. With this information, it is 
possible to design the thermopile which will give optimum 
response under the obtaining conditions. 

The characteristic sensitivity of a thermopile determines 
its response and, in the theory, this quantity Q is denned as 
follows : 

Q = ^P- (2) 

$ is the radiant energy falling on the receivers in unit time, 
I is the current in the galvanometer-thermopile circuit, and 
R is the total resistance in this circuit. Q is in effect like 
an efficiency — the efficiency with which the radiant energy 
to be measured is converted into galvanometer deflections. 
The expression for Q for an uncompensated vacuum ther- 
mopile of n junctions in terms of the quantities on which it 
depends is 

27 Cartwright, C. H., Zeits.f. Physik, 92, 153 (1934). 


H $ y/Ri + R. + R.l 

where I is the thermoelectric current in the thermopile- 
galvanometer circuit, R is the total electrical resistance 
of the circuit, made up of the thermopile resistance R h 
the galvanometer resistance R g , and any external resist- 
ance R e . P is the combined thermoelectric power of the 
thermoelectric wires, expressed in volts per degree centi- 
grade, a is the Stef an-Boltzmann radiation constant, A 
the area of the receiver, T the absolute temperature of the 
receiver, and e its effective radiating power. Wi and W 2 
are the Wiedemann-Franz coefficients of the two thermo- 
couple wires. 

The quantity in the brackets represents the total heat 
losses of the receiver. The middle term in the brackets 
represents heat loss by conduction through the wires, and the 
third term represents heat loss due to the Peltier effect. 
Ordinarily the influence of Peltier heat on the design may be 

The first term in the brackets represents the heat lost by 
radiation and gas conduction. Where the receiver is not in 
a high vacuum, gas conduction has the same effect on ther- 
mopile design as increasing the magnitude of e and, as we 
have pointed out before, the use of a receiver with a small 
emissivity for heat radiation, e, has the effect on thermopile 
design of decreasing the quantity eA. 

Fig. 21 illustrates for a vacuum thermopile the way in 
which Q depends on the values of eA, the number of junc- 
tions, and the total electrical resistance in the thermopile 
circuit. With e taken as unity the curves are constructed 
for A = 1 mm 2 and A = 3 mm 2 . Furthermore, these 
curves are for thermoelectric wires made of pure bismuth 
and wires of bismuth plus 5 per cent tin having a thermo- 
electric power of 120 microvolts/ °C. and Wiedemann-Franz 



[Chap. VIII 

coefficients of 3 X 10" 8 watt ohm/°C. 2 and 4.2 X 10" 8 watt 
ohm/°C. 2 respectively. The full curves are for thermopiles 
having one, two, three, and four junctions, and the dotted 


O lO 20 30 40 50 6l> 10 80 90 

J{x in ohms 

Fig. 21. 

curves are for compensated thermopiles having one and two 
active and compensating junctions respectively. 

It is desirable, from a practical point of view, to have a 
minimum number of junctions to build. The information 
given in Fig. 21 facilitates making the compromise between 


this practical consideration, on the one hand, and the desire 
to have a maximum sensitivity on the other. From curves 
in this figure, it is apparent that the energy should be con- 
centrated onto a receiver which is as small as possible. 

By reference to Eq. 3 we see that when the third term in 
the brackets is small in comparison with the first and second 
terms, the sensitivity, Q, appears to be proportional to the 
thermoelectric power, P. This is not always the case in 
practice, and a thermoelectric metal should not be chosen 
on the basis of the thermoelectric power alone. As a matter 
of fact, most metals with a high thermoelectric power have an 
unfavorable Wiedemann-Franz coefficient, which may, in the 
end, make them even less desirable than metals such as the 
bismuth alloys, which are convenient to manage. 28 

Sensitivity and minimum energy detectable. When the 
quantity Q, given by Eq. 2, is combined with the current 
sensitivity, dd/dl, and the total resistance of the circuit, R, 
it yields the composite sensitivity, S, of a thermopile and 
critically damped galvanometer according to the formula 


Here 8 is the deflection of the galvanometer caused by the 
radiant energy 3> falling on the receiver in unit time. 

It has been customary to compare the sensitivities, S, of 
the various radiometric instruments. This has led to some 
confusion in the literature. Actually, in making the most 
delicate radiometric measurements, we are not interested 
primarily in the value of S (which can be made as large as 
desired by the use of an amplifier) but rather in the accuracy 
with which the radiant energy can be measured in a given 

28 Cartwright, C. H., Ann. d. Physik, 18, 656 (1933). The Wiedemann- 
Franz coefficient, W, of any metal can be determined by using the empirical 

W = 2.32 X 1(T 8 + 3 X lO- 1 ^ watt ohm/°C. 2 , 

where p is the specific electrical resistivity and T the absolute temperature. 
For good conductors p is small, so that W is the same for all these substances. 


time, or, what amounts to the same thing, in the smallest 
intensity of radiant energy that can be measured in a given 
time with a given accuracy. The magnitude of this smallest 
deflection is influenced by disturbances acting on the in- 

We will designate this smallest deflection that can be 
measured by a single reading in a time t , and with a mean 
relative error g, by the symbol min . Until 1926 it was 
considered that the elimination of the disturbances on which 
the value of min . depends was simply a matter of refining 
experimental technique. Ising was the first to point out 
that our experimental technique is already advanced far 
enough so that in many cases min . is determined by the 
ever-present Brownian motion fluctuations. 29 If we 
consider the thermopile system isolated from all disturb- 
ances except those produced by Brownian motion of the 
galvanometer, then the value of min . is easy to determine. 
According to the principle of the equipartition of energy, 
every object with one degree of freedom, such as the moving 
system of our galvanometer, will possess a definite amount of 
kinetic and potential energy. The average value of the 
kinetic energy or potential energy at 19 °C. is 

ikT = 2 X 10" 21 watt sec. (5) 

The average deflection due to the potential energy is in- 
volved in the expression 

Potential energy = \K& = ikT, (6) 

where K is the torsional constant of the suspension and k is 
Boltzmann's constant. When a reading is taken, the 
fluctuations of 6 give rise to an uncertainty amounting to 
■\/k T I K, Therefore, in order to have a probable error of g, 
a single deflection must be at least 1/g times the average 
fluctuation, or 

»_-!,§. (7, 

29 Ising. G. t Phil Mag., 1, 827 (1926). 


It can be shown that this expression is a general one appli- 
cable to any radiometric instrument. Combining Eq. 7 
with Eq. 4, we get an expression for the least energy that can 
be measured : 6 6 X 10- 11 

*-• = ' SgVk Watt - (8) 

In comparing the $ min . of different radiometric instru- 
ments, it is necessary to specify not only the accuracy factor 
g, but also the time t, to be taken for measuring a deflection. 
In the case of a galvanometer, this is because the value of 
dd/dl depends on t . The value of S also depends on t for 
other radiometric instruments. It is not correct to assume, as 
is usually done, that the value of <£ min . varies with the square 
of the period of the deflecting device. As a matter of fact, 
in the case of a thermopile and critically damped galva- 
nometer, the value of <£ min . is proportional to the square root 
of the period time of the galvanometer. 30 

The 3> min . of a thermopile and galvanometer can be 
expressed in terms of the factor g, the Q of the thermopile, 
and the period of the galvanometer, as follows: 

1.1 X 10- 


**■ = gQVu watt (9) 

With the values of Q given by the curves in Fig. 12 it is 
therefore possible to estimate the minimum energy falling on 
the receiver in unit time that can be measured with a pro- 
posed apparatus. It is to be observed that the sensitivity of 
the galvanometer does not enter Eq. 9. Eq. 9, however, 
does imply that the deflections are measured either directly 
or with the help of an amplifying device to the limit set by 
Brownian motion. 

General summary of the work on thermopile design. The 
remainder of this chapter will be devoted to a summary of 
the results of experimental and theoretical investigations 
made by one of the authors, C. Hawley Cartwright, on the 

30 Cartwright, C. H., Physics, 1, 211 (1931). 
Czerny, M., Ann. d. Physik, 12, 993 (1932). 


relative merits of the different radiometric instruments, and 
in addition will present some general (although not neces- 
sarily final) conclusions resulting from these studies. 

Vacuum microradiometers can be made which will measure 
less energy, ^ min ., than the best vacuum thermopiles used 
with a separate galvanometer. This advantage is not 
sufficient to offset the practical advantage of greater flexi- 
bility of the thermopile with separate galvanometer. 

Vacuum bolometers and vacuum thermopiles have at 
present practically the same limit, , <£ min . If a material 
with better characteristics than nickel were available for the 
construction of the bolometer strip, this situation would 
be altered. 

Radiometers will not respond to as small energies, $ min ., 
as thermopiles. The direct comparisons made by the 
author, especially in Berlin and Brussels, between vacuum 
thermopiles and radiometers yield results in favor of vacuum 
thermopiles. Radiometers are usually much more sensitive 
than thermopile and galvanometer combinations, owing to 
the use of a much lighter moving system than is possible 
with a galvanometer. Brownian motions are, however, 
increased, so that they more than offset the advantage of 
the larger primary deflections. 

A question of considerable importance and one which bears 
on the above conclusions is the following: Why is there 
often considerable variation in the sensitivity of vacuum 
thermopiles, in fact, sufficiently large variations to be re- 
sponsible for many of the publications that have appeared 
on improving thermopiles? The answer is that many 
vacuum thermopiles are not constructed with the maximum 
possible sensitivity, for the following reasons : 

1. The sensitivity of a thermopile depends on the skill 
exercised in its construction. 

2. For the most part, thermopiles have been constructed 
without first calculating the proper design or, if this is done, 
without dependable information on the physical properties 
of the materials used. 


3. A sufficiently high vacuum is not always used. A 
properly designed and constructed thermopile should be 
about twenty times more sensitive in high vacuum than in 
air, and, on increasing the vacuum from 10~ 3 to 10~ 6 mm of 
mercury, the sensitivity should be doubled. 

4. The thermoelectric power of the bismuth and bismuth- 
alloy wires is often less than 120 microvolts/ °C. Slight 
impurities can greatly influence the thermoelectric power of 
bismuth by influence on crystal orientation, and so forth. 
For example, the thermoelectric power of pure bismuth 
relative to copper changes from 57 to 107.7 microvolts/ °C. 
for different crystal orientations. 31 

5. The influence of deviations of the properties of bismuth, 
and especially bismuth alloys, from the predictions of the 
Wiedemann-Franz law is generally neglected, with the result 
that thermoelectric wires with a resistance which is too small 
are used so that the sensitivity falls on the left-hand steep 
part of the curves corresponding to those shown in Fig. 21. 

Actually, the ultimate attainable sensitivity for a thermo- 
pile is limited by the unfavorable departure from the Wiede- 
mann-Franz law of the thermoelectric metals that possess a 
high thermoelectric power. However, if this were not the 
case, it is interesting to note that the thermoelectric power 
itself would limit the sensitivity. From Eq. 3 we see that 
for a thermoelectric power of 250 microvolts/ °C. the heat 
loss due to the Peltier effect is equal to the heat loss due to 
conduction through the wires. Although the possibility 
exists of finding better thermoelectric metals than bismuth 
and the alloy of bismuth and 5 per cent tin, it seems rather 
improbable that much progress will be made in this direction. 

It is well to keep in mind that although tin has ten times 
less specific electrical resistance than bismuth, an alloy of 
bismuth and 5 per cent tin has twice the specific electrical 
resistance of pure bismuth. This should be considered when 
better thermoelectric metals are being sought. Bismuth 

31 Bridgman, P. W., Am. Acad., Proc, 63, 347 (1927-1928). 


itself is an unusually favorable metal for thermopiles, not 
only because it has a relatively high thermoelectric power, 
but also because it is a pure metal element having a small 
specific electrical resistance and does not depart greatly 
from the Wiedemann-Franz law. 

In order to improve the sensitivity of thermopiles, there 
is the possibility of using them at low temperatures, where 
Q can be increased, owing to a greater thermoelectric power, 
a more favorable Wiedemann-Franz ratio, and less radiation 
loss from the receivers. However, liquid-air thermopiles 
have several practical disadvantages. 32 

32 Cartwright, C. H., Rev. Sci. Instruments, 4, 382 (1933). 


Optics: Light Sources, Filters, and 
Optical Instruments 

Divisions of the spectrum. The electromagnetic spectrum 
divides naturally into the region for which the eye is sensitive, 
the infrared region, with frequencies below those which we 
perceive as red, and the ultraviolet region, with frequencies 
higher than those which we perceive as violet. These 
regions are defined roughly by the wave lengths given in 
Table I. In the text we will use microns for expressing wave 
length in the infrared and Angstroms for expressing wave 
length in the visible and ultraviolet. The visible region 
includes less than one octave of frequency, while the so-called 
infrared region embraces at least nine octaves and the ultra- 
violet embraces five or six octaves. 

Light sources. The sun. The sun naturally comes first 
in consideration of light sources. Its use is recommended 


wave lengths in Angstrom units 


He NaNa 

KH hg GG'd 


1 \ 


bjb 2 b 4 E 

I / 

HU atmoar 

C ph6 g 

Fig. 1. 

for many experiments because of its brightness and because 
in the Fraunhofer lines it contains numerous convenient 
wave-length landmarks. The Fraunhofer lines, which are 
conspicuous in the spectrum exhibited by a good pocket 
spectroscope, are shown in Fig. 1. 




[Chap. IX 

The energy distribution in the solar spectrum, as observed 
through the atmosphere, is closely approximated by that 
of a black body at 5400 °K. The luminous efficiency of the 

O 1 ooo 2ooo 3ooo 4ooo 5ooo 6000 7ooo Sooo 9 000 lo,ooo 
temperature in absolute degrees 

Fig. 2. 

sun is about 80 lumens/ watt. As will be seen in Fig. 2, this 
is nearly as high an efficiency as it is possible to attain with a 
heated body. 


Divisions of the Electromagnetic Spectrum 

Spectral Region 

Wave-Length Limits 

Extreme ultraviolet 

500 A to 2000 A 






2000 A to 4000 A 
4000 A to 4460 A 
4460 A to 4640 A 
4640 A to 5000 A 

Near in 






5000 A to 5780 A 
5780 A to 5920 A 
5920 A to 6200 A 
6200 A to 7200 A 
0.72^ to 20m 

Far inf: 

^diate infrared 

20m to 40m 


40m to 400m 

A heliostat or coelostat is required if a beam of sunlight is 
to be maintained in a fixed direction in the laboratory. 
Heliostats are obtainable from scientific supply companies. 

Chap. IX] 



Their mirrors, which are usually silvered on the back, should 
be recoated on the front surface with aluminum if it is 
desired to obtain in the reflected sunlight the full range of 
solar spectrum down to the atmospheric cutoff at approxi- 
mately 3000 A. 

The details of construction for a home-made coelostat are 
shown in Figs. 3 and 4. This coelostat may be driven by 

J summer sun 
h* 2nd mirror 

/equinoctial son 

y^wintct" sun 

1^4 mirror between 
these guides in 

T^^-rS l&* mirror between these 
^3 Y guides in afternoons 

ist rnirror 
in summer 

1 st mirror 
at equinox 

l^t mirror 
in winter 

2.^ mirror may be directed down 
through a port in the roof if this 
arrangement is preferred. 

Fig. 3. 

shifting of the 
\^t mirror from 
one set of guides 
to the other is to 
avoid the shadow 
of the 2^ mirror 
near noon. 

the works of an alarm clock as shown; it may also be driven 
by a Telechron clock. The secondary mirror of the coelostat 
has controls operated by cords for making adjustments. 

Tungsten lamps. Tungsten lamps are the most con- 
venient laboratory source of white light. Their efficiency is 
about 11 lumens/ watt for the nitrogen-coiled filament type. 

The differences of spectral energy distribution of various 
tungsten-filament lamps are illustrated in Fig. 6, Chapter 
XI. The spectrum of emission of the filament is limited in 
the ultraviolet and infrared by the transmission of glass. 
With glass bulbs } mm in thickness, the spectrum extends 
from about 3100 A in the ultraviolet to 3/z in the infrared. 



[Chap. IX 


polar axis should 
lie in the plane 
of the 



fork and pin drive 
from setting pm 
of clock 

scale of inches 

5 AO 


detail of 1^ mirror mounting 

Fig. 4. 

Two tungsten lights convenient for many purposes in the 
laboratory are shown in Fig. 5. The one shown on the left is 
a projection lamp. It requires 6 volts and 18 amperes. An 
autotransformer or high-capacity storage battery serves as 
power source. The battery is, of course, preferred when con- 
stancy and steadiness of the emission are important. 1 The 

1 The autotransformer is as satisfactory as the battery when it is energized 
bv the output of a Raytheon voltage regulator. 

Chap. IXj 



4 volts 

scale of inches 
Fig. 5. 

lamp shown at the right has a straight filament. It is 
useful as a galvanometer lamp. Both of these lamps are 
obtainable commercially. 2 

A trade-mark on the end of 
a tungsten lamp bulb, when it 
interferes with the light emis- 
sion of the filament, may be 
removed by polishing with 
rouge and felt or with wet 
crocus cloth. 

A lamp 3 with a quartz bulb 
for absorption spectra is 
shown in Fig. 6. The bulb 
contains argon at 1J atmos- 
pheres pressure. The tung- 
sten operates at about 31 00 °C. 
and gives a continuous emission spectrum extending into the 
ultraviolet to 2500 A. At the operating temperature, the va- 
por pressure of tungsten is appreciable, and it would normally 

blacken the quartz part of 
the bulb. However, vertical 
convection currents of argon 
gas carry the evaporated 
tungsten molecules upward 
from the filament, so that 
they are not deposited on the 
quartz but rather on the up- 
per glass part of the bulb, 
where they do not impair the 
usefulness of the lamp. 

Welsbach mantle* This re- 
fractory mantle was formerly 
used extensively for house 

JO amperes 
-tungsten at 

about 30-mH 
tungsten wire 

continuous spectrum to 250uA 
Fig. 6. 

2 These lamps may be obtained from the General Electric Company, Nela 
Park, Cleveland, Ohio. 

3 This lamp is supplied by the Philips Laboratory, Eindhoven, Holland. 

4 Ives, H. E., Kingsbury, E. F., and Karrer, E., "A Physical Study of the 
Welsbach Mantle," Frank. Inst, J., 186, 401, 585 (1918). 

346 OPTICS [Chap. IX 

illumination and is now used in gasoline lamps. It is brought 
to incandescence in the outer hot surface zone of a Bunsen 
burner type of flame, where it assumes a temperature nearly as 
high as the Bunsen flame temperature. The mantle is com- 
posed of thorium oxide with 0.75 to 2.5 per cent cerium oxide 
added to increase its visible emissivity. This addition of ceri- 
um oxide plays much the same role as the sensitizer for a 
photographic plate ; that is, it introduces an absorption band 
in a desired spectral region without materially affecting the 
optical properties elsewhere. The effect of the cerium oxide is 
to make the emission in the green 30 percent greater than that 
of a black body at 1800 °C, whereas the emissions in the red 
and blue correspond closely to 1800°C. color temperature. 5 
The near infrared emissivity is less than 1 per cent from 0.7 /jl 
to about 6/x, and the incapacity of the mantle to radiate heat in 
this important region accounts for its high temperature. For 
the spectrum beyond 10/i the mantle again has an emissivity 
greater than 75 per cent. The mantle is an excellent labora- 
tory source for those long wave-length infrared radiations/ 

Barnes suggests heating the mantle with a sharp oxygen 
flame striking it at a grazing angle. 7 This gives it a higher 
temperature, and also the elongated heated section produced 
is properly shaped for illuminating the slit of a spectrometer. 
More recently, Pfund has devised an arrangement to com- 
bine both electric and flame heating, allowing the attainment 
of even higher temperatures. 8 

Nernst glower. Nernst filaments are composed of zir- 
conium dioxide powder with about 15 per cent yttrium 
oxide powder. 9 For operation on alternating current, flex- 
ible platinum lead wires are later cemented to each end 

* Forsythe, W. E., /. 0. S. A., 7, 1115 (1923). 

6 Rubens, H., Deutsch. Phys. Gesell, Verh., 7, 346 (1905); Ann. d. Physik, 
18, 725, (1905), 20, 593 (1906); Phys. Zeits., 6, 790 (1905), 7, 186 (1909). 

7 Barnes, R. B., Rev. Sci. Instruments, 5, 237 (1934). 
» Pfund, A. H., /. O. S. A., 26, 439 (1936). 

9 Nernst, W., and Bose, E., Phys. Zeits., 1, 289 (1900). 
Nernst glowers are obtainable from StupakorT Laboratories, 6627 Ham- 
ilton Avenue, Pittsburgh, Pennsylvania. 

Chap. IX] 



of the refractory tube with a mixture of the oxide powders 
and zirconium chloride as a binder. For operation on direct 
current, the manner of attaching the electrodes is more 
complicated. The Nernst lamp normally operates at around 
2000 °K. Its spectrum extends well into the ultraviolet and 
infrared. However, beyond 15/* its emission is said to be 
inferior to the emission of the Globar heater. 

At one time the Nernst glower offered great promise for 
commercial lighting, owing to a luminous efficiency of 6 
lumens/watt as compared with 3 lumens/watt for the carbon 
filament. However, the modern incandescent lamp with a 
coiled tungsten filament in an atmosphere of nitrogen, having 
an efficiency of 11 lumens/watt, entirely changed matters. 
The use of the Nernst light is now confined to the laboratory. 
Here its usefulness depends upon the fact that it is operated 
in air and has a convenient form (cylinder 0.4 to 0.6 mm in 
diameter and 1 to 2 cm long) for focusing on the slit of a 
spectrometer. Griffith has 
described details of construc- 
tion for making Nernst fila- 
ments. 10 

Since the Nernst lamp has 
a negative temperature co- 
efficient of resistance, it must 
be stabilized with external re- 
sistance or, better, with a bal- 
last lamp having an iron-wire 
filament mounted in hydro- 
gen. 11 The iron wire of this 
lamp runs at a faint red glow; 
its remarkable current-stabilizing effect in an atmosphere of 
hydrogen at 30 to 100 millimeters pressure is shown in Fig. 7. 
Such a ballast lamp consumes 10 or 15 per cent of the total 
power needed for operating the Nernst filament. 

10 Griffith, H. D., Phil Mag., VI, 50, 263 (1925). 

11 For the theory of the hydrogen ballast lamp, see Busch, H., Ann. d. Physik, 
64, 401 (1921). 

voltage across lamp 
arbitrary units 

Fig. 7. 



[Chap. IX 

Globars. The Globar is a rod of bonded silicon carbide 
about -TQ inch in diameter and about 10 inches long". The 
ends fit into aluminum cup electrodes. A potential of 100 
volts across the rod brings it to an orange or yellow heat. It 
can be operated in air at a temperature above 1000 °C, 

although at temperatures 




plunger *• 

rod - Me 
in diam- 
eter -lo" 
long * — 

bra 35 

around 2000°C. the carbide 
dissociates and carbon is va- 
porized or oxidized, leaving 
silicon, or, in the presence of 
air, silicon dioxide. A protec- 
tive layer of thorium dioxide 
sintered to the outside of the 
Globar with thorium chlo- 
ride as binder will allow of 
temperatures in excess of 
2000°C. 12 A suitable mount- 
ing for the Globar is shown 
in Fig. 8. 

Carbon arcs. The carbon arc 
is useful as a laboratory light 
source. Ordinarily, the posi- 
tive carbon is mounted hori- 
zontally. An 8-mm positive 
carbon is consumed at the 
same rate as a 6-mm vertical 
negative carbon. Accordingly, 
if carbons of this size are used, they may be fed into the arc 
automatically by clockwork. 

The carbon arc requires at least 40 volts to operate it. 
Higher voltage increases the size of the positive crater with- 
out materially affecting its surface temperature. 

The character of the light emission from the ordinary 
carbon arc may be influenced by the addition of metallic 
salts as cores in the carbons. (Magnesium fluoride is often 
used to get a white arc.) The spectral distribution of the 

12 1 am indebted to C. H. Cartwright for this information. 

openings to 



" tubing 


Fig. 8. 

Chap. IX] 



carbon arc with cored carbons is illustrated in Fig. 9. This 
is a curve of galvanometer deflections against wave length, 
as determined with a quartz monochromator (shown in 
Fig. 32) and cesium oxide photocell. (See Chapter X.) 
The slit widths were the same for all wave lengths. This 
curve does not correct for the transmission of the image- 
forming lens (shown in Fig. 5, Chapter XI) which was used 
to focus the light. Without this lens the spectrum would 
have extended well into the ultraviolet. 




X t* 



I s 


.2 o 

-H o 

**- 2 

^' 2 

2 mm Bol lamp focused on 
monochromator slit with a 
projection lens 
2.6 amperes 

positive crater of carbon arc 
focused on monochromator 
slit with a projection lens 
8 mm carbons, 5amperes 

o (J 

o u 

w rt3 



o <-> 


3000 4000 5000 6000 7000 8000 9000 10.000 11.000 

wave length in Angstrom units 
Fig. 9. 

The ordinary carbon arc has a crater brightness of about 
13,000 candles/cm 2 and an efficiency of about 35 lumens/ 
watt. The Sperry Gyroscope Company has produced an 
arc that uses special shields to confine the current to a definite 
boundary around the rotating crater. 13 This arc is about 
six times as bright as the ordinary arc. 

Lummer has succeeded in obtaining extreme temperatures 
in the carbon arc by operating it in an inert atmosphere 

13 Benford, F., Trans. Soc. Motion Picture Eng., 84, 71 (1926). 



[Chap. IX 

A1 arc** 

under high pressure. Under a pressure of 22 atmospheres 
he was able to obtain temperatures of 7600 °K., considerably 
in excess of the solar surface temperature. The surface 
brightness reported for this temperature was 280,000 can- 
dles/cm 2 . The attainment of such temperatures and bright- 
ness is difficult. 

A technique of measuring. For the preliminary study of a 
spectrum plate, a technique of measuring and recording data 

which is neat and avoids con- 
fusion is illustrated in Fig. 10. 
This procedure employs an 
enlarged print of the original 
spectrum plate to identify the 
iron or other reference lines 
appearing in the eyepiece of 
the comparator. To facilitate 
this identification, the wave 
lengths of the iron lines are 
written in the margin of the 
print. Also, the print serves 
as a permanent record of the 
appearance of the spectrum as 
well as a record of the data of 

First, the wave lengths of 
conspicuous iron comparison 
lines, which are to be used as reference lines in the measure- 
ment, are written in the margin. The original plate has the 
same appearance in the eyepiece of the comparator as the 
enlarged print; thus it easily serves to identify the com- 
parison lines. After the wave length of each unknown line 
is determined by interpolation, it is recorded on the clear 
margin of the print as shown at the left in Fig. 10. Notes 
may also be added in this margin when the wave lengths are 
later identified by reference to Kayser's tables. 14 

14 Kayser, H., Tabelle der Hauptlinien der Linienspektra alter Elemente. 
Berlin: Julius Springer, 1926. 



O6.0 33 

Fig. 10. 

Chap. IX] 



Iron arcs. The iron arc is used in the laboratory by the 
spectroscopist as a source of ultraviolet light and also as a 
standard comparison source. Its spectrum has been thor- 
oughly studied, and the wave lengths of the lines, as well as 
the influence of pole and pressure effects on them, are well 
known. 15 

An iron arc developed by Pfund 16 suitable for use in the 
laboratory is shown in Fig. 11. An iron oxide bead is 
placed on the lower electrode for stabilizing the arc. If the 

bush in 

iron 12 mm 

rods of 


Fig. 11. 

upper electrode is a graphite rod, the arc is even more stable 
than it is with an iron electrode. 17 The arc can be started 
by rubbing a carbon across the gap. 

Low-pressure mercury arcs. The low-pressure mercury 
arc is a convenient laboratory light source. 18 It gives several 
strong lines in the visible, ultraviolet, and near infrared 

15 See the following: 

Babcock, Harold D., Astrophys. J., 66, 256 (1927), 67, 240 (1928). 
St. John, Chas. E., and Babcock, Harold D., Astrophys. J., 46, 138 (1917), 
58, 260 (1921). 

16 Pfund, A. H., Astrophys. J., 27, 298 (1908). 

17 The National Carbon Company produces a spectroscopic grade of pure 
graphite. The pure carbon arc exhibits only one line in the visible or the 
ultraviolet spectrum. This line is 2478 A. 

18 For a description of a simple, home-made, low-pressure arc. see Pfund, A, 
H., Asirophys. J., 27, 299 (1908). . 

352 OPTICS [Chap. IX 

spectra. These lines are far enough apart to be separated 
with filters. (See Table XI.) 

The ultraviolet spectrum of the arc in a fused quartz tube 
extends to about 2000 A. The energy at the extreme short 
wave-length limit produces ozone in the air. The ozone 
formation, however, becomes weaker and weaker as the lamp 
is burned, owing to changes in the transmission limit of the 
quartz. Finally the ozone formation practically ceases. 
Baly has found that such changed quartz will emit a green 
phosphorescence and will regain its original transparency if 
it is heated in the blast burner. 19 

The Cooper-Hewitt type of mercury light has a brightness 
of about 2.3 candles/cm 2 . The ordinary Cooper-Hewitt 
illuminating lamp has a tube 4 feet long and about 1 inch in 
diameter. It is a convenient light source for many experi- 
ments when an extended source is desired, as for observing 
Haidinger's and Newton's fringes. To get uniform illumi- 
nation over an extended area, drafting linen is hung below 
the lamp. 

In glass the Cooper-Hewitt lamp does not, of course, emit 
all of the ultraviolet spectrum. Recently this arc has been 
put on the market, made with a tube of Corex red-purple 
glass which suppresses the visible radiation (except 4046) 
and transmits the near ultraviolet. In this form it is excel- 
lent for therapeutic use. 

The commercial hot quartz vacuum arc is much more 
brilliant (350 candles/cm 2 ) than the Cooper-Hewitt lamp 
discussed above. The ordinary hot quartz lamp is not of a 
convenient form for use in the laboratory, but it is now avail- 
able in the form of a vertical straight quartz tube constructed 
especially for laboratory use. 20 These laboratory arcs are 
equipped with rectifiers, so that they may be operated on 
either alternating or direct current. 

19 Baly, E. C. C, Spectroscopy. New York: Longmans, Green and Com- 
pany, 1927. 

10 This lamp and the one discussed above are obtainable from the Cooper- 
Hewitt Electric Company, Hoboken, New Jersey. 

Chap. IX] 



High-pressure mercury arcs. Harries and Hippel 21 have 
described a high-pressure mercury lamp which is now com- 
mercially available. 22 This is illustrated by Fig. 12. The 
lamp is mounted in a nearly light-tight case — a very conven- 
ient construction for use in the laboratory. The lamp is 
made of uviol glass or quartz, with or without added cad- 
mium to obtain the red cadmium 6438 A line. Schott glass 
filters are also supplied for isolating the yellow, green, blue, 
violet, or ultraviolet lines. 

The spectrum of the high-pressure lamp exhibits con- 
siderable continuous background. Accordingly, the spectral 

resistance coil 
in counterpoise 
adjustable for 
11© or 220 volts 

purity obtainable with it by the use of niters is not as great 
as it is with the low-pressure arc. The emission, however, is 
very steady, especially when the lamp is operated on storage 

Cornelius Bol of Stanford University (formerly of the 
Philips Laboratory, Eindhoven, Holland) has developed a 
so-called super-high-pressure mercury arc. 23 The discharge 

21 Harries, W., and Hippel, A. v., Phys. Zeits., 83, 81 (1932). 

22 This lamp is obtainable from Schott und Gen., Jena, Germany. Their 
agent in this country is Fish-Schurman Corporation, 250 East 43rd Street, 
New York. 

23 Bol, C., Das Licht, 5, 84 (1935); Ingenieur, 50, 91 (1935). 
Barnes, B. T., and Forsythe, W. E., /. 0. S. A., 27, 83 (1937). 
Dushman, S., J.O.S.A., 27, 1 (1937). A bibliography of high-efficiency 

light sources is given 



[Chap. IX 

which produces the high pressure is started, however, by 
argon at a pressure of 2 or 3 cm of mercury. The operating 
potential for the lamp is around 500 volts. Heat generated 
by the argon discharge volatilizes the liquid mercury exposed 
in the lamp until a pressure of mercury gas of about 200 
atmospheres is attained. On account of the high ultimate 
pressure, the lamp must be made of a thick- walled capillary 
tube as shown in Fig. 13. The tungsten electrodes project 
beyond the reserve mercury in order to guide the discharge 
down the central part of the tube. In the center, tempera- 
tures of 8600 °C. and brightness values several times greater 

x /z mm 


quartz appendix sealed 
off leaving the proper 
amount of mercury 
to allow electrodes to 
project Vx mm 

quartz capillary 
6mm outside 

30- or40-mil 
e I c ex rode 

The whole lamp is enclosed 
in a glass water jacket. 

Fig. 13. 

than the brightness of molten tungsten are attained. For 
example, a lamp operating on 640 volts at a pressure of 
200 atmospheres has a brightness of 180,000 candles/cm 2 
and a luminous efficiency of 79 lumens/watt. The emission 
of a Bol lamp is shown in Fig. 9. (See also Table II.) 

The inside surface of the quartz capillary probably attains 
a temperature in excess of the critical temperature of mer- 
cury, so that no liquid mercury can condense. The mercury- 
gas envelope around the hot central core of the arc absorbs 
the resonance line emitted in the core, and at the obtaining 
pressure and temperature the resonance line is so broad that 
its absorption extends over the major part of the ultraviolet 
spectrum (to 2700 A). 

The electrodes are sealed in the Bol lamp with a new 

Chap. IX] OPTICS 355 

glass. A lamp of convenient size for use in the laboratory 
has the electrodes spaced 1 cm apart. It is first filled with 
2 cm pressure of gaseous argon and then with liquid mercury 
until the 30-mil tungsten wires project about \ mm beyond 
the mercury at each end. A 640- volt transformer is suitable 
for operating the light. It is connected in series with the arc 
and a suitable choke coil. When the arc is shorted out, the 
choke will draw about 3.4 amperes from the transformer. 24 

A "cold," low-pressure mercury-vapor lamp is shown in 
Fig. 14. 25 This lamp employs a few millimeters pressure of 
hydrogen, argon, or one of the other noble gases as a starting 
gas. Heat developed by the discharge in the noble gas soon 
distills mercury vapor from small globules of the liquid metal. 

.^quartz, tube with a longitudinal partition 

~7 r ' fSM^^^VVS^ 

opening through partition ^tungsten electrodes 

Fig. 14. 

The potential for operating the lamp is obtained from a sign 
transformer or from a storage battery and spark coil. This 
lamp is only about one tenth as brilliant in the visible as the 
Harries and Hippel lamp, but its emission at 2536 A is many- 
fold greater. In fact, about 80 per cent of its total emission 
is in the resonance line. 

The resonance fine from the mercury lamp shown in 
Fig. 14 is so strong that the mercury vapors, rising from a 
globule of liquid mercury held in the hand, cast a strong 
shadow on a fluorescent screen. 26 With a 3-mm Corex red- 
purple filter to suppress the visible spectrum, this lamp is 
ideal for exciting the fluorescence of minerals. 

This type of mercury light is very useful in the laboratory. 

24 The Bol lamp must be operated surrounded with a stream of cool- 
ing water. 

25 This lamp is obtainable from the Reed and Miller Company, 16 South 
Raymond Street, Pasadena, California. 

26 See Leighton, W. G., and Leighton, P. A., Jour, of Chem. Ed., 12, 139 



[Chap. IX 

When neon is used instead of argon as the starting gas, this 
single source yields a series of strong lines well distributed 
over the spectral range from 2536 A to 10,140 A. The gap 
in the mercury spectrum between 6907 A and 10,140 A is 
filled by a series of neon lines around 8300 A. 27 

Filters for use with the various mercury arcs to yield 
monochromatic light are discussed in a later section. 

Other gaseous discharges. Commercial sodium arcs are 
now available. They are confined in a special glass con- 
tainer that is not attacked by the metal vapor. 28 These arcs 
operate inside a clear Dewar flask and afford a large-area 
source of monochromatic light which is particularly suited 
to many laboratory tests and demonstrations. The char- 
acteristics of this and the Bol lamp are given in Table II. 


Characteristics of Super-High-Pressure Mercury Lamp and Sodium 



Sodium Lamp 

Mercury Lamp 

Pressure (atmospheres) .... 

Current density 

Cross section (cm 2 ) 

Candles/cm 2 

Vapor temperature (°C.) . . . 
Light output (lumens/watt) 

10" 5 
10 to 20 




1.8 X 10 6 



Heller, G., Philips Techn. Rev., 1, 2 (1936). 

Mercury lamp: 1400 watts and water cooled, 1.3 amperes, 2 mm cross sec- 

Sodium lamp: 100 watts in a clear Dewar flask. 

Pyrex is not attacked by sodium as readily as are soft 
glasses, and by fusing borax or boric acid to the inside sur- 
face, its resistance to the alkali metal can be further in- 

27 For wide monochromator slits, the tungsten lamp is a much richer light 
source in this region than the argon discharge. 

28 Buttolph, L. J., Am. Ilium. Eng. Soc, Trans., 30, 147 (1935). For similar 
lamps using other metallic vapors, see Alterthum, H., and Reger, M., Das 
Licht, 8, 69 (1933). 

Chap. IX] 



creased. 29 Magnesia crystals are not attacked by vapors of 
the alkali metal, and they may be used for experiments in 
which sodium, at higher temperatures and pressures, is to be 
confined behind windows transparent to both the ultraviolet 
spectrum and the infrared spectrum. 30 

The ultraviolet spectrum obtained from a hydrogen dis- 
charge tube is continuous, extending from the short wave- 
length emission limit of incandescent tungsten toward shorter 
wave lengths to the limit of transmission of quartz. This 
hydrogen continuum is most effectively excited by sources of 
the type developed by Duffendack and Manley, Smith and 
Fowler, Munch, and Jacobi. 31 These sources excite the 
spectrum with thermoelectrons emitted from a hot cathode. 

Capillary discharge tubes 
filled with many different 
elementary gases are now 
available commercially. 32 

Sparks. To obtain the 
spark spectrum characteris- 
tic of the materials com- 
posing the electrodes, it is 
necessary to use a condenser 
of sufficient capacity to 
give an explosively noisy 
spark. Either a transformer or an induction coil can be 
used as the source of potential. A spark between mag- 
nesium electrodes, especially if it is confined between glass 


Of light 




Fig. 15. 

29 See Chapter XIV. 

30 Brice, R. T\, Rev. Sci. Instruments, 8, 209 (1937). 
Strong, J., and Brice, R. T., J. O. S. A., 25, 207 (1935). 

31 Duffendack, O. S., and Manley, J. H., J. 0. S. A., 24, 222 (1934). 
Duffendack, O. S., and Thomson, K. B., J. 0. S. A., 23, 101 (1933). 
Herzberg, G., Ann. d. Physik, 84, 553 (1926). 

Jacobi, G., Zeits.f. techn. Physik, 17, 382 (1936). 

Lau, E., and Reichenheim, O., Zeits.f. Physik, 73, 31 (1931). 

Lawrence, E. O., and Edlefsen, N. E., Rev. Sci. Instruments, 1, 45 (1930). 

Munch, R. H., Am. Chem. Soc., J., 57, 1863 (1935). 

Smith, A. E., and Fowler, R. D., /. 0. S. A., 26, 79 (1936). 

32 These tubes may be obtained from the Central Scientific Company, 
Chicago, Illinois, and A. D. Mackay, 198 Broadway, New York City. 



[Chap. IX 

A concentrated 
solution of the 
salt being studied 
is placed in the 
container. Dilute 
HCI and pieces 
of zinc are added. 
Bubbles rising 
form &. spray 
which is drawn 
into the burner, 
^coloring the 

plates, is very brilliant. Such a light source, shown in 
Fig. 15, is useful for shadow photographs of bullets in motion, 
and so forth, and for the photography of sound waves by the 
Schlieren-methode. 33 The duration of the illumination from 
the magnesium spark can be made extremely short. 

Flames. Flames such as the Bunsen flame, which are 
almost colorless, give characteristic emission spectra when 
volatile metallic vapors are introduced. The metals most 
commonly used to obtain monochromatic or nearly mono- 
chromatic light are given in 
Table III. 

Sodium light may be ob- 
tained by wrapping asbes- 
tos, soaked in sodium chlo- 
ride, around the tip of the 
Bunsen burner tube. An- 
other method of introducing 
the salts into the flame is il- 
lustrated by the device shown 
in Fig. 16. A neodymium fil- 
ter may be used to absorb the 
emission of sodium vapor and 
at the same time transmit the 
red emission lines from po- 
tassium or lithium vapors. To obtain the metallic thallium 
spectrum, a bead of the metal, fused in a platinum- wire loop, 
is touched to the edge of the Bunsen flame. The bead is 
introduced just far enough to obtain the desired rate of evapo- 
ration. If the bead is held too far inside the flame, it boils 
away rapidly. Inasmuch as thallium is a poisonous metal, a 
high concentration of the vapors in the room should not be 
allowed. Also, sodium, potassium, and lithium vapors may 
be introduced into a Meker burner flame by placing a small 
globule of fused sodium chloride, potassium chloride, or 
lithium chloride on the grill of the Meker burner. 

33 Wood, R. W., Physical Optics, page 93. New York: The Macmillan 
Company, 1934. 

Fig. 16. 

Chap. IX] OPTICS 359 

The ultraviolet. The portion of the ultraviolet spectrum 
treated here will be limited to the wave-length range 2000 A 
to 4000 A. 34 In the long-wave half of this region between 
4000 A and 3000 A many substances are transparent, includ- 
ing mica, celluloid, diamond, Canada balsam, ether, glycerin, 


Flame Spectra 


Wave Length of Emission Lines 


5890 A, 5896 A 


7665 A, 7699 A 


6708 A 


5350 A 

acetone, turpentine, xylene, and in thin layers, many ordi- 
nary glasses. (See Table IV.) For the entire region from 
4000 A to 2000 A the list of materials is not so great. It 
includes rock salt, potassium chloride, fluorite, magnesia, 
lithium fluoride, alum, gypsum, sugar, calc-spar, water, 
ethyl alcohol, glacial acetic acid, liquid ammonia, fused and 
crystalline quartz, and cellophane. (For the transmission of 
cellophane see Table V.) 

Prisms, lenses, and mirrors for the ultraviolet. Only 
a few of the substances mentioned above are suitable for 
making prisms and lenses. Fluorite and quartz make excel- 
lent prisms. They can be combined to make achromatic 
lenses. But the scarcity of fluorite of good optical quality in 
large sizes makes these achromats very expensive. A combi- 
nation of quartz and rock salt is sometimes used for making 
achromats. Recently, synthetic alkali halides and mag- 
nesium oxide have become available in large pieces, and 
these, together with other synthetic substances, will no 
doubt become important for constructing ultraviolet optics. 

34 For a general treatment of ultraviolet radiations, see Luckiesch, M., 
Holladay, L. L., and Taylor, A. H., Frank. Inst., J., 196, 353 (1923). 



[Chap. IX 

The optical constants of some of these materials for the 

visible spectrum are given in Table VI. 

Concave aluminized mirrors are now used for ultraviolet 

optical systems. They have the same focus for ultraviolet 

as for visible light, and therefore they can be adjusted 



Transmissions of Glass (5 mm Thickness) in the Ultraviolet 

Bausch and Lomb 





Wave Length 

Condenser Glass 



































Spectral Characteristics of Cellophane — Per Cent Transmission at 

Indicated Line 

Standard Colorless Cellophane 














P. T 

















Standard Plain-Colored Cellophane 































Dark green .... 
Dark blue 



This table was supplied by E. I. duPont de Nemours and Company, Wil- 
lineton, Delaware. 

mington, Delaware. 

Chap. IX] 



Index of Refraction of Synthetic Materials 
















































Fused quartz 

CaF 2 






Plexiglas. . . . 










Filters for the ultraviolet. Thin metal films are among 
the most interesting filters for the ultraviolet. The trans- 
mission band exhibited by silver and the alkali metals is 
associated with a gap lying between the region where the 
reflection is ascribed to the effect of free electrons (on the 
long wave-length side of the gap), and the region where 
reflection is ascribed to bound electrons (on the short wave- 
length side). In silver, this gap at 3160 A is approximately 
100 A wide. It is much wider than this for the alkali metal 

Potassium films may be used as a filter for isolating ultra- 
violet radiations. The full transmission of potassium in the 
ultraviolet begins at 3000 A for films of a thickness just 
sufficient to be opaque in the visible to sunlight. R. W. 
Wood has studied this phenomenon and describes how these 
films can be formed on a quartz-glass bulb cooled to liquid 
air temperatures. 35 Unfortunately, films prepared as he 
describes are only permanent at temperatures considerably 
below room temperature. O'Bryan, however, has shown 
how potassium may be deposited between quartz-glass plates 
to give films permanent even at the elevated temperature of 

36 Wood, R. W., Phys. Rev., U, 353 (1933). 



[Chap. IX 

boiling water. 36 The transmission of these thicker films 
begins at about 3350 A, becomes about 25 per cent at 
2500 A, and decreases to a little below this value as the wave 
length 2000 A is approached. The transmission of such a 
potassium film is illustrated by Fig. 17. 

zinc spark 

zinc sparK with 
potassium f i rter 
3 000 4 00O 500 O' 

wave length in Angstrom units 

Fig. 17. Transmission of a potassium film. 36 

Bromine vapor can also be used as a filter. It is trans- 
parent to the ultraviolet rays. A layer of saturated bromine 
vapor 5 cm thick at room temperature is opaque to blue 
light and nearly opaque to green light, as one can readily see 
by interposing a bottle containing a little liquid bromine 





3 000 


no filter 
acetic acid 
tartaric acid 

acetic acid 


mercuric chloride 

% /a 

ph e ny laceticacid 




5ooo The fractions 

wave length in Angstrom units above represent 

the amounts of 
an approximately 
saturated solu- 
tion which is 
added to a unit 
of water. 

Fig. 18. Transmission spectra of various materials. After Williamson, R. C, 
Phys. Rev., 21, 111 (1923). 

between a mercury lamp and a pocket spectroscope. The 
ultraviolet transmission of bromine begins at 3800 A, and 
the vapor is quite transparent to the spectrum from wave 
length 3500 A down to at least 2345 A. 

36 O'Bryan, H. M., Rev. Set. Instruments, 6, 328 (1935). 

Chap. IX] 



A 5-mm layer of a solution of nitrosodimethylanalin 
(10 mg to 100 cc water) has about the same transparency as 
the bromine vapor. 37 

A filter of 14 g pure, iron-free nickel sulphate crystals and 
10 g pure cobalt sulphate crystals dissolved in 100 cc dis- 
tilled water is opaque to the visible spectrum but transparent 

carbon disulphide 

benzyl alcohol 




ethyl methyl Ketone 

amyl alcohol 


ethyl oxalate 

normal butyl alcohol 

ethyl benzoate 






butyl acetate 

ethyl propionate 

carbon tetrachloride 

ethyl formate 

ethyl acetate 

formic acid 

amyl acetate 

acetic acid 

iso propyl alcohol 




methyl alcohol 

ethyl aJcohbl 

2000 2500 3000 3500 4000 

wave length in Angstrom units 

Fig. 19. Transmission of various liquids. After Brode,W. R., /. Phys. Chem.. 

30, 56 (1926). 

in the ultraviolet below 3300 A. In layers 3 cm thick this 
filter transmits 3.5 per cent of the 3342 A mercury line and 
96 per cent of 3126 A line, and it is transparent as far down in 
the ultraviolet as 2300 A. 38 

37 Wood, R. W., Phil. Mag., 5, 257 (1903). 

38 Backstrom, H. L. J., Naturwiss., 21, 251 (1933). 

364 OPTICS [Chap. IX 

The ultraviolet transmission limit for mica is at about 
2800 A for 0.01 mm thickness. Mica of this thickness is 
completely opaque at wave lengths below 2600 A. 

The transmissions in the ultraviolet of some other ma- 
terials are illustrated in Figs. 18 and 19. 

Polarization of the ultraviolet. The new sheet polarizers 39 
made of herapathite are opaque to ultraviolet light. (See 
Fig. 38.) Although the calcite of Nicol prisms is trans- 
parent to 2000 A, the Canada balsam used for cementing 
them is not transparent in the ultraviolet at wave lengths 
below about 3000 A. For cementing optical surfaces to be 
used in the ultraviolet, glycerin, castor oil, or dextrose sugar 
should be used. A Wollaston prism may be used to polarize 
light in the ultraviolet when its parts are properly cemented. 

The infrared. The infrared spectrum extends from 
7600 A, or 0.76^1, to about 400/z. A thermopile or radiom- 
eter is generally used for measuring infrared radiation. As 
the operation of these instruments depends on thermal 
effects produced by the radiation, the infrared spectrum is 
often referred to as the heat spectrum. The infrared radia- 
tions are emitted by heated bodies. Ordinarily, heated 
bodies are used as laboratory sources for the infrared spec- 

It is convenient to divide the heat spectrum into three 
regions: The near infrared, from 1.1/* to 20 jjl; the inter- 
mediate infrared, from 20/x to 40 jjl; and the far infrared, 
from 40/z to 400/x. The spectroscopic significance of the 
near infrared is that the characteristic frequencies of gases 
which fall in this region generally arise from molecular 
oscillations, whereas the characteristic frequencies which fall 
in the visible and ultraviolet regions arise in general from 
electronic oscillations. On the other hand, in the far infra- 
red the characteristic frequencies of gases arise from mo- 
lecular rotation and molecular bending. In the case of 
crystals the characteristic frequencies in the near infrared 

39 Land, E. H., Frank. Inst, J., 22!>, 269 (1937). 
Freundlich, H., Chemistry and Industry, 56 } 698 (1937). 

Chap. IX] OPTICS 365 

are generally interatomic oscillations within the chemical 
radicals that exist as units in the crystal, while frequencies 
in the far infrared are due to oscillations of the positive ions 
(or radicals) of the crystals relative to the negative ones. 

The intermediate infrared spectral region from 20/x to 40ju 
was formerly closed to investigation on account of the lack of 
transparent substances to be used for making windows and 
prisms. There are now available, however, a transparent 
paraffin of high melting point, 40 and large synthetic crystals 
of the alkali halides which are transparent in the range 20/j 
to 40/x. 41 

Prisms, windows, lenses, and mirrors for the infrared. 
The important prism materials for the infrared are listed in 
Table VII. These materials are not ordinarily combined to 


Transmission of Materials for Infrared Radiations 


Useful Transmission Limit 
in the Infrared 



CaF 2 








form achromatic lenses for focusing the infrared rays; 
mirrors which are much more satisfactory are used. Even 
spherical mirrors are useful for the less exacting work, since 
the slits in infrared spectroscopy can never be set as fine as 

40 Kellner, L., geb. Sperling, Zeits.f. Physik, 56, 215 (1929). The paraffin 
in question is Kurlbaum, M. P., 68° to 72 °C. 

41 Bridgman, P. W., Am. Acad., Proc, 60, 307 (1925), 64, 19 (1929). 

Korth, K., Zeits.f. Physik, 84, 677 (1933). 

Kyropoulos, S., Zeit. f. anorg. allgem. Chem., 154, 308 (1926). 

Ramsperger, H., and Melvin, E. H., J. 0. S. A., 15, 359 (1927). 

Stober, F., Zeits.f. Krist., 61, 299 (1925). 

Strong, J., Phys. Rev., 86, 1663 (1930). 



[Chap. IX 

J 2 

*> o 

£ * 

s * 

o o 

-1 J. ^v^.1 . 


— ' >>vw * t— jrj — i — - 

V+ >«■ 4 + 4 

0.5 to l^u thicK 




4 \y^/ 4- 4 14 j 

+ f + 4 4 

1 + 

15m thicK +~ — *" 

+• ' + 4 +- 

4 1 


4/ 1 4 4 / V 4 

/ 4 

\7 \ 


1 1 i — i — i 



1 1 I l i 

' 4 



Ol 1 1 1 —J ' 1 1 

1 ^ 3 4 5 6 7 8 9 lO 11 1% 13 14 15 

wave length in yu 
Fig. 20. After Barnes, R. B., and Bonner, L. G., J.O.S.A., 26, 433 (1936). 

they can in the other spectral regions, in which photography 
can be applied.- 2 

Materials useful for windows on absorption cells and 
vacuum radiometric devices are listed in Table I, Chapter 
VIII. (See also Figs. 20, 21, 22.) Of these materials the 
high-melting-point paraffin is of special interest, since it is 
one of the few materials opaque to the near infrared spectrum 
and transparent to the long wave lengths. Soot is another 
such material. Although it is quite opaque in the visible, 
soot is translucent for the heat spectrum. 

The reflection of most metals such as silver, speculum, and 
aluminum is high in the infrared. The reflectivity for wave 




E 9 

5 o 

O. o 

paraffin (melting point 68° to 70* C) 

2. lmm 

4 5 6 T 5 9 10 11 

wave length in jul 

13 14 t5 

Fig. 21. After Barnes, R. B., and Bonner, L. G., J.O.S.A., 26, 433 (1936). 

42 Strong, J., Phys. Rev., 37, 1661 (1931). 

Chap. IX] 



reflection of 
quartz jjlate 

fused quartz 
O.6o5mm thick 

transmission of 
owdered qua 

4 5 6 7 

wave length in u, 


Fig. 22. Infrared transmission and reflection of quartz. After A. H. Pfund. 

lengths longer than about 10^i can be calculated from the 
electrical conductivity of the metal by the expression 

#x = 1 - 0.365VpA, (1) 

where p is in ohms mm 2 /m and X is in microns. 

Reflection of crystals. Residual rays. Crystals exhibit 
so-called bands of "metallic" reflection at certain wave lengths 
where the reflection coefficient, usually of the order of 5 per 
cent, approaches 100 per cent. This property of crystals was 
first observed by E. F. Nichols. 43 The bands of high reflec- 
tivity exhibited by quartz, for example, are shown in Fig. 23. 

reflection of 
crystalline quartz. 

+ + + 

* + + 

7 S 9 lO 12 14 1618 20 25 30 40 50 60 70 8O90100 150 200 250 300 

wave lengfth in^u 
Fig. 23. After H. Rubens. 

43> Nichols, E. F., Ann. d. Physik, 60, 401 (1897); Phys. Rev., L, 297 (1897). 
Rubens, H., and Nichols, E. F., Ann. d. Physik, 60, 418 (1897): Phys. Rev., A, 
314(1897). . 



[Chap. IX 

Quartz (for the ordinary ray) exhibits two strong bands, one 
at 8.9jtt and one at 20.8/x. Rock salt has only one band, 
at 52^t. 

Multiple reflections from crystals are employed to isolate 
narrow bands of monochromatic radiation from the heat 
spectrum. For example, if the spectrum from a heated 
body is reflected once from a rock-salt crystal surface, the 
energy at wave lengths about 52/x are reflected while those 
radiations elsewhere, especially in the short-wave spectrum, 
where the reflection is nonmetallic, are attenuated about 


Number of 




























(3 mm paraffin 
in each case) 

1 cm KC1 
5 mm KC1 

3 mm KBr 


0.4 mm quartz 
1.2 mm KBr 

0.4 mm quartz 

2 mm quartz 
2 mm quartz 
2 mm quartz 
2 mm quartz 
2 mm quartz 
2 mm quartz 

Wave Length 


















Energy (cm 
of deflection; 
scale at 3 m) 












Chap. IX] 



twenty times. In spite of this attenuation by a single 
reflection, the energy in the 52/x band may still be much less 
than the integrated energy reflected at other wave lengths. 
After a second reflection, however, the short-wave spectrum 
is again attenuated about twenty times, or four hundred 
times altogether, while the energy in the band of waves 
around 52/* is little affected. Accordingly, after four or 
five reflections the only radiations remaining, the so-called 
residual rays, are those of the 52/jl band. 

The use of these successive reflections is a standard pro- 
cedure for obtaining monochromatic bands of radiation in 
the far infrared. The crystals used for obtaining various 
wave lengths are listed in Table VIII. We shall describe 
the apparatus used for obtaining residual rays in a later part 
of this chapter. 

Special absorbers for the near infrared. Water is trans- 
parent from wave lengths greater than 0.2/x in the ultra- 
violet throughout the visible spectrum. (See Fig. 24.) 

c *" 




H o 



ZVx% solution of + 
copper chloride 

% cm thick 

water 1cm thicK 

+ + + 

500 600 700 800 900 1000 llOO ROO 1300 1400 1500 1600 

wave length in nryu 
Fig. 24. 

However, it is opaque in the heat spectrum for all rays be- 
yond the limits X for thickness r , as given in Table IX. 

A water filter is often used to absorb the heat rays that are 
emitted when a carbon arc, the sun, or a tungsten lamp is 
used as a light source. The use of a water filter prevents 
the cracking of lantern slides with heat, burning of photo- 



[Chap. IX 


Transmission Limits of Water in the Infrared for the 
Solar Spectrum 


Xo in \i 

1 mm 


1 cm 


10 cm 


10 m 


100 m 


Fowle, F. E., Smithsonian Miscellaneous Collections, 68, 49 (1917). 
Schmidt, W., Meterolog. Zeitsch., 25, 321 (1908). 

graphic film, overheating of microscope objectives, or exces- 
sive heating of polarizing Nicols. 

The addition of cupric salts to water results in increased 
absorption of the infrared. The absorption for the infrared 
is illustrated in Fig. 24 for a 2-cm cell containing cupric 
chloride. 44 

Manufactured glass filters such as Aklo glass and the 
Schott filters BG17 and BG19 are designed to remove the heat 


Transmission of Aklo Heat-Resistinq Glass (2 mm Thickness) for the 
Light of a Vacuum Mazda Lamp (2360°K.) 








Extra Light Shade No. 395 






Light Shade No. 396 


Medium Shade No. 397 


Dark Shade No. 398 


Glass Color Filters, Corning Glass Works, Corning, New York. 

44 Absorption of water: Nicholson, Seth B., and Pettit, Edison, Astrophys. J., 
56, 295 (1922). 

Absorption of cupric chloride solution: Coblentz, W. W., Bureau of Stand- 
ards Scientific Paper No. 168. 



4 mm 

4 mm 





3 000 

46oo 5ooo 6000 

wave length in Angstrom units 
W=Wratten gelatin filters G and Noviol are Corning glass filter^ 

4,000 5 000 6 000 7 000 

wave length in Angstrom units 

Fig. 25. Transmission of glass and Wratten filters. 




[Chap. IX 

spectrum. 45 (See Table X and also Jena Colored Optical 
Filter Glasses, obtainable from Fish-Schurman Corporation, 
250 East 43rd Street, New York City.) The transmission of 



land 3.5 mm. 


Tfccl Purple 
Cores. A 
3 m 






0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 

wave length in yd 

Fig. 26. From Glass Color Filters, Corning Glass Works, Corning, New York. 

BG17, 1 mm thick, and BG19, 4 mm thick, is about the same as 

that of 2 cm of a nearly saturated copper sulphate solution. 

Visible spectrum. Glass and gelatin niters are used for 

isolation of the mercury lines. They are easier to handle 

and much more permanent 

index of refraction 
of the components 
h of a filter + 


than water solutions. The 
transmissions of some of the 
glass and gelatin filters com- 
mercially available in this 
country are illustrated in 
Figs. 25 and 26. A list of 
the filter combinations for 
the separation of various 
spectrum lines is given in 
Table XI. 

The Christiansen filter. The 
Christiansen filter consists of 
a mass of solid particles im- 
mersed in a liquid medium, as, 

46 Heat-absorbing glass is manufactured by the Corning Glass Company, 
Corning, New York. BG17 and BG19, manufactured by Schott und Gen., 
are handled in this country by the Fish-Schurman Company, New York City. 

4ooo n 5ooo 6000 7ooo 
Angstrom units 

Fig. 27. After McAlister. 
footnote 46.) 


Chap. IX] 



for example, particles of borosilicate glass immersed in carbon 
disulphide and benzene. 46 Fig. 27 shows the dispersion for a 
borosilicate crown glass and for a 10 per cent solution (by 

glass windows 



pinhole light" 

In large filters 
metal vanes should 
be inserted to help 
maintain a uniform 
temperature through- 
out the contents. 

d gasket 
ubber gasket 


^ ^ '"dTaphragm 

filter of powdered 
glass and a hquid- 
of the same refrac- 
tive index. 

approximately \ 



Fig. 28. After McAlister. 46 

volume) of carbon disulphide in benzene (both anhydrous) 
at 20 °C. The filter composed of these two transmits freely 
the color for which the indices of refraction of the liquid and 
solid phases are identical, that is, where the two lines in 
Fig. 27 cross. For this color the medium is optically homo- 
geneous. The filter is a nonhomogeneous optical medium 
for all other wave lengths. 
Accordingly, they are scat- 
tered. By means of the ar- 
rangement shown in Fig. 28, 
the scattered waves are iso- 
lated from the freely trans- 
mitted color. The individual 
transmissions of five filters 
are shown in Fig. 29. These 
filters were 18 mm thick and 
were made up from borosili- 
cate glass using different 
concentrations of carbon di- 
sulphide in benzene. 


transmission of 

a set of five filters 


C * 

+ + 4- » - 




1 "" 1 " 


\\ ! \ 


\\ 1 i 

► 1 1+ t ■ 



1 1 1 1 1 

8 2 

• h» / \ h \ /+ V 



OOO 5 OOO 6 000 7 OOO 

Angstrom units 


ng. : 

29. After M 

cAlister. 46 

46 Christiansen, C, Ann. Physik u. Chemie, 23, 298 (1884), 24, 439 (1885). 
McAlister. E. D., Smithsonian Misc. Coll., 93, No. 7 (1935). 



[Chap. IX 

One limitation of the Christiansen filter lies in its lack of 
complete opacity to wave lengths on either side of the trans- 
mitted band. 

Filters for Isolating Mercury Lines 






Infrared or 

G554EK, 6 to 8mm 

88 as used by 
R. W. Wood or 

Cobalt blue glass and satu- 
rated solution of potas- 
sium dichromate 

5769 to 

G34R, 3 to 4 mm 

22 Hg yellow 

and eosin 



G34Y, 3 to 4 mm 

62 Hg green or 
77 Hg special or 
77 A. Hg special for 

Neodymium ammonium 
nitrate and potassium di- 


Noviol A, 3 mm, and 
G585, 3 to 5 mm 

50 Hg blue 

Cobalt blue glass 
and quinine sulphate 

4047 to 

G586A, 3 to 5 mm, 

Noviol 0, 3 to 4 mm 

36 Hg violet 

Methyl violet and 
quinine sulphate 


8 to 10 mm 

18 ultraviolet 

Methyl violet and 
acid green 

Buttolph, L. J., Engineering Bulletin 104-B, Cooper-Hewitt Company. 

This limitation is a serious one. For example, when the 
filter is to be used in conjunction with a highly selective 
receiver, such as a photocell, the response of the receiver for 
rays weakly transmitted by the filter but for which the 
receiver is exceptionally sensitive (or for which the emission 
of the source is especially strong) may seriously interfere 
with the interpretation of the results obtained. Another 
limitation of this filter is its sensitivity to temperature 
changes. The filter cannot be used effectively in an intense 
beam of light such as sunlight, owing to temperature gradi- 
ents set up in the cell. 

However, the dependence of the transmitted wave length 
upon temperature may be put to use. F. Weigert and 

Chap. IX] 



collaborators have found, for example, that a cell made of 
particles of crown glass immersed in liquid methyl benzoate 
transmitted red light at 18°C. (64°F.) and blue light at 50°C. 
(122°F.). 47 

A very interesting Christiansen filter effect is exhibited by 
the infrared transmission of thin powder films. 48 Their 
maximum of transmission occurs at the wave length at 
which the index of the powder is unity or equal to the index 
of the surrounding medium. For magnesia this transmission 
maximum in air is at 12.2/x, and if the filter is immersed in 
carbon tetrachloride, the maximum shifts to wave length 9/z, 
at which both the carbon tetrachloride and the magnesia 
have the same index. 

Reflection of metals. Of the metals useful in the visible 
spectrum for reflection of light, the three most important 
are aluminum, speculum, and silver. Their reflectivities are 
shown in Fig. 30. It is to be noted that aluminum is 












_ 4- 











I ~ 4- 

. 4- 




i 4. 





/ Stellite*-^ 






< Speculum. 






> 50 





















O 40 
pi 30 


\ ■*■ 1 
































. . 1 


1 — 


. , 1 . 


_ 1 


— « 

aooo 3000 4ooo 5ooo 6000 7ooo Sooo Qooo 10,000 11,000 12,000 

wave length in Angstrom units 
Fig. 30. 

47 Weigert, F., Staude, H., Elvegard, E., and Shidei, J., Zeits.f. phys. Chem., 
AH. B, 2, 149 (1923), 9, 329 (1930). 

"Barnes, R. B , and Bonner, L. G., Phys. Rev., 49, 732 (1936). 



[Chap. IX 

superior to newly deposited silver for all wave lengths less 
than 4100 A. In the visible spectrum the use of aluminum 
instead of silver is recommended. Although new silver has a 
better reflectivity in the visible spectrum than aluminum, it 
soon tarnishes. 

The apparatus shown in Fig. 31 was used for the above 
reflectivity measurements. This apparatus measures the 

mirror comparison ft / \ 


mirror to 
be tested is 
inserted in 
this frame 



from light 


screws to comparison 

make the mirrors 

three mirror 

frames parallel 

and equally 


'to photo- 



'pull screw 

mirrors held in 
place with 
paper clips 

Fig. 31. 

from light 



square of the absolute reflectivity directly (putting the 
comparison mirror in both the numerator and denominator, 
so to speak). 

Monochromators. The best method of isolating a narrow 
wave-length band of high spectral purity from a source of 
white light is to use a double monochromator, that is, two 
single monochromators built together. High spectral purity 
is often desirable for highly selective effects, such as, for 
example, the determination of the long wave-length limit of 
the photoelectric effect, or in any other case when the slight 
spectral impurity that one might have with a single mono- 
chromator would vitiate the results of the measurement. A 

Chap. IX] 



step can be made in the direction of high spectral purity by 
the use of niters in series with a single monochromator. 
These filters are, however, usually less efficient than a single 
monochromator. The transmission of a single monochrom- 
ator is about 45 per cent. 

The monochromator may have achromatic lenses, but 
these are very expensive if they are constructed of materials 

quartz prisms 
alumintzed on bach 

screw to main- 
tain focus through 
out the .spectrum ^ 

drum calibrated 
in wave lengths 


quartz prisms with 

front faces ground to 

form lenses 

quartz reflecting 
prisms um 

1 •t'-^JL //mediate 

leveling screw 

A partition (not shown) 
prevents stray light from 
one system entering 
the other. 

This cam sliding on the pin 
rotates the rear prism table 
maintaining minimum devia- 
tion throughout the spectrum 

AH the slits are separately 
adjustable. The entrance and 
exit slits are curved to 
compensate for prismatic 

Fig. 32. Hilger-Miiller double monochromator. 

which will function in the ultraviolet. Generally, mono- 
chromators employ quartz lenses. These are brought to 
focus with a mechanism operated by the wave-length drum. 
Fig. 32 shows how this is accomplished in the Hilger-Mtiller 
double monochromator by the use of a cam bar mounted on 
the prism table. As the prism table and lens system move 
as a unit toward the slit system, the lenses are brought into 
focus for shorter and shorter wave lengths. The cam bar is 



[Chap. IX 

so constructed that it causes the wave lengths to fall on the 
exit slits for which the lenses are in focus. 

Use of mirrors in monochromators. Parabolic mirrors are 
often used in monochromators, because an optical system 
using mirrors is achromatic. However, mirrors have the 
distinct disadvantage as compared with lenses that the 
parallel collimated beam is returned in the direction of the 
entrance slit, a direction which precludes a neat simple 
arrangement of the other optical parts. To use a mirror on 

slit source of li 

perforated j 
flat / 




'/ ) 




off- ax is 




of lig 









lb) * 

of light >' 

parallel lidht 

Fig. 33. 

its optical axis requires either an auxiliary flat as in the 
Pfund 49 arrangement shown in Fig. 33(a) or an off-axis 
mirror as shown in Fig. 33(b). One way to make such an 
off-axis mirror is to construct a large ordinary paraboloidal 
mirror and cut out the desired mirror from one side of it. 

A mirror system composed of spherical mirrors like the 
one shown in Fig. 33(c) may be used. This, of course, 
introduces large distortions in the wave front. It is possible, 
however, by proper orientation of a similarly imperfect 

49 Pfund, A. H., J. 0. S. A., 14, 337 (1927). For a grating spectrometer 
application of Pfund's scheme see Randall, H. M., Rev. Sci. Instruments, S ( 
196 (1932). 

Hardy. J. D., Phys. Rev., 38, 2162 (1931). 

Chap. IX] 



mirror to compensate in a measure for the distortions pro- 
duced by the collimator and to obtain better definition than 
would be possible even if a perfect telescope were used. The 
proper arrangement of the telescope system for achieving 
this compensation is shown in Fig. 34, with the regular 
Wadsworth arrangement. 50 

plane mirror 

parabolic mirror 
prism ^ 

parabolic mirror 

• exit slit or 

Fig. 34. 

entrance slit 

The optical train in monochromators is usually either the 
Littrow arrangement or the Wadsworth arrangement, both 
of which use the prism at minimum deviation. These 
arrangements are shown in Fig. 35. 51 

Water monochromator. An ultraviolet monochromator 
with improvised optics, devised by Harrison, 52 is shown in 
Fig. 36. The optical parts consist of a water prism and a 
spherical aluminized mirror. This monochromator is very 
simple, and optically it is good enough for isolating the 
stronger mercury lines (as the illustration of the produced 
spectrum shows). It has a relatively high aperture, //6. 
The dispersions of crystal quartz, fused quartz, and water 
are related as 25:21:19 at 3000 A. Since water is more 
transparent to the ultraviolet than quartz, this mono- 
chromator can be used for isolating wave lengths as short as 
1820 A. 

50 Czemy, M., and Turner, A. F., Zeits. f. Physik, 61, 792 (1930). 
Czerny, M., and Plettig, V., Zeits. f. Physik, 63, 590 (1930). 
81 Littrow, O., Am. J. Set., 85, 413 (1862). 

Wadsworth, F. L. O., Phil. Mag., 88, 137 (1894); Astrophys. J., 2, 264 
62 Harrison, George R., Rev. Sci. Instruments, 5, 149 (1934). 



[Chap. IX 

plane mirror 

axis of rotation 
of prism and 

and exit slits 
are placed one 
over the other 

normal Wads worth 

The least-deviated 
ray emerges paral- 
lel to the entering 
ray but displaced 
laterally an amount 
which remains con- 
stant re Oard I ess 
of the rotation of 
the system. 

modifiecKWads worth arrangements 

The least-deviated 
ray em ergfes at 

of the system. 

Fig. 35. 

The least-deviated 
ray emergres at 
right angles to the 
entering ray but 
shifts literally 
with the rotation 
of the system. 

Focal isolation. Fig. 37 shows the method of focal isola- 
tion invented by Wood to isolate the far infrared radiations 
from a Welsbach burner. 53 When the first lens is positioned 
in relation to the light source at a distance equal to twice its 
focal length for the far infrared rays, where the index of 
refraction is 2.25, the near infrared rays emerging from the 
lens are divergent. An opaque spot at the center of the 
quartz lens prevents the direct transmission of the median 
near infrared rays through the aperture provided at the focus 
of the far infrared rays. Usually two lenses are arranged 
in series of effect complete separation of the far infrared 

A focal isolation method has been applied to the isolation 

Rubens, H., and Wood, R. W., Phil. Mag., 21, 249 (1911). 

Chap. IXj 



Exit slit ^ v 
should be 
cut to con- 
form to the 
shape of 
the images. 




Exit-slit frame should be 
hinged on a horizontal 
axis permitting adjustment 
to a sharp focus throughout 

the spectrum. 

Interior of instrument 
should be painted black 

tubes for 

distilled water 

earthenware bowl 

spherical mirror 
coated with aluminum, 
platinum, or chromium 

If aluminum is used, the 
water should be drained 
when the instrument is 
not in use, otherwise the 
mirror may peel. 

tube for 
filling and 

leveling screws. 
may be used for fine 



2804 3125 3650 4358 

wave length in Angfstrdm units 
Fig. 36. 

'A drawing from a 
photograph of part 
of the mercury-arc 
spectrum taKen with 
Ian instrument similar 
I to that above, but 
using an oblique si it 



[Chap. IX 

of the 1940 A group of aluminum lines with a quartz lens. 54 
And, while there is for quartz no such diversity of index in 
this part of the spectrum as there is in the infrared, yet these 
lines are separated from the rest of the aluminum spectrum 
with a spectral purity of 0.98. The intensity obtained is 
sevenfold greater than that obtainable from a quartz mono- 
chromator. This focal isolation method has also been 
applied to the 2030 A to 2140 A group of zinc lines. 

first stage 
quartz lens 

second stage 
quartz \cns 

disk to 
central cone 
of short- 
wave radiation 

Welsbach mantle 
radiation source 

loo/x mean 
vvave length of 
isolated radiation 

Fig. 37. 

Residual-ray isolation. Apparatus using the residual-ray 
method for the isolation of wave lengths in the infrared is 
illustrated in Fig 38. 55 The apparatus shown at the top of this 
figure employs four crystal reflections, while the one at the 
bottom, placed at the focus of an image-forming mirror, uses 
only two crystal reflections. 

When the two-crystal apparatus is equipped for 6.7ju 
(crystals of calcite) it is useful for measuring humidity, since 
this region of the spectrum is very sensitive to moisture in the 
radiation path. On the other hand, with either quartz, 
Carborundum, or potassium chromate crystals, which give 
bands of radiation at 8.7/x, 12/x, and 11.6/z, respectively, the 
instrument is useful as a radiation pyrometer insensitive 
both to water vapor and to light smoke or haze. In the 
region from 8/x to 13/* there is very little absorption by the 

54 Forbes, Geo. S., Heidt, Lawrence J., and Spooner, Lawrence W., Rev. Sci. 
Instruments, 5, 253 (1934). 
65 Strong, J., Phys. Rev., 87, 1565 (1931), 88, 1818 (1931). 

Chap. IX] 



A g as U My 

absorption/ X 
cell may ' /\ 
be placed 

aluminum -coated 
spherical mirv-or 

detail of crystal holder 

to galva- 


That radiation 
should be used 
which is emitted 

almost tangentially- not shown 

The case is built of brass and 
sealed with wax so the air inside 
may be Kept dry. The case should 
be insulated with felt/ which is 


parallel light 
from source 
being studied 


spherical mirror 

Fig. 38. 

two -junction 
vacuum ther- 
so that a 
slight rotation 
of the carrier 
brinos the „ 
alternately to 
the focus 



[Chap. IX 

water in the air even when it is humid; in this region of the 
spectrum the entire thickness of the atmosphere exhibits a 
transmission comparable to the transmission of the atmos- 
phere for green and yellow light (T = 85 per cent). 

Polarization. There are now new polarizers available for 
use in the visible spectrum, but they are not as efficient as 
Nicol prisms. 56 The transmission of these polarizers, shown 
in Fig. 39, does not yield as high efficiency as that of a Nicol 
prism. For plane-polarized light of proper azimuth a Nicol 
prism transmits about 80 per cent. Two Nicol prisms in 

c » 

O O 
if) o 

& q 

Marks Polarizing Plates 

+ + +■ » + -+ 

parallel *. 

i. ^ + 

$ 3 
8 * 

J- o 
3 000 

4ooo 5ooo 6ooo 7ooo 8000 9 000 
wave length in Angstrom units 

Fig. 39. 

series transmit a maximum of about 32 per cent of un- 
polarized white light. At the other extreme, two Nicol 
prisms accurately crossed are quite opaque. For example, 
they will not transmit enough sunlight to make the disk of 
the sun discernible. However, to obtain this degree of 
opacity, the Nicol prisms must be crossed very precisely (to 
an accuracy of the order of 1 second of arc). 

The new polarizers have the advantage over Nicol prisms 
that they can polarize a beam of greater aperture (both areal 
and angular). Two applications of the new polarizers are 
illustrated in Figs. 40 and 41. 

One of these, illustrated in Fig. 40, applies to the measure- 
ment of strain in glass. Objects to be tested for strain, as, 

66 Strong, J. ; J. 0. S. A., 26, 256 (1936). 

Chap. IX] 



for example, glass-to-metal seals, are immersed in a jar fitted 
with parallel glass sides and containing a liquid medium 
having the same index of refraction as the glass. This 
medium may, for example, be a mixture of the proper pro- 
portions of carbon disulphide and benzene or a mixture of 
zylene and alcohol. Polarized light obtained from a lamp by 
reflection off black glass at the polarizing angle (or reflection 
off the back of an exposed photographic plate which has been 
developed, fixed, and dried) is viewed through a full-wave 

plane-aided glass tank 

for the immersion of 

irregular objeets in lamp housin 

a liquid of equal 

refractive index *. condensing lens 

condensing" lens 
negative lens 


ca one 
wave plate 

black glass or 
plate glass 
painted black on 
under side 

analyzer -Nicol prism 

If Polaroid or a similar polarizing" material is 

used, the negative lens may be omitted. 

Fig. 40. 

mica plate and analyzer. (The construction of the full-wave 
plate is described below.) When a full-wave plate is placed 
in front of the analyzer, slight variations of the polarization 
over the field of view are manifest as variations of color from 
the purple of the unstressed condition. 

Engineering applications of polarized light. The property 
of isotropic transparent materials that a strain makes them 
double refracting is used by engineers for studying the mag- 
nitude and distribution of stress produced by loading various 
two-dimensional structures, such as, for example, the shapes 
represented by the cross section of a dam. 57 An arrange- 

67 Brahtz, J. H. A., Rev. Sci. Instruments, 5, 80 (1934). 
Goetz, A., Rev. Sci. Instruments, 5, 84 (1934). 



[Chap. IX 

ment for such studies using spherical mirrors and the new 
polarizers is shown in Fig. 41. The astigmatism (due to 
using the mirrors off axis) can be balanced out, at least in 
part, by tipping the camera lens about a horizontal axis by a 
suitable amount. The model of the shape to be tested is 
usually made from a clear sheet of Bakelite or Marblette. 
Table XII gives the coefficient of forced double refraction 
for various materials suitable for constructing models. 

specimen being tested 

lenses " 

' 'I 

Polaroid disks 

light source. 

objective lens 
tilted to correct avtigr 
screen or 
photographic plate 

Fig. 41. 

The quarter-wave plates are used in the illustrated ar- 
rangement to allow the elimination of the pattern of iso- 
clinics (the lines along which the principal stress in the 
specimen has a constant inclination) from the pattern of 
isochromatics (the lines along which the quantity (p — q) 
has a constant value). Here p and q are the principal 
stresses produced in the model by the applied loading. 
Methods of determining the magnitude of the quantities 
p and q from the measured isoclinics and isochromatics can- 
not be described here, since they are quite complicated. 58 

68 Coker, E. G., and Filon, L. N. G., A Treatise on Photo-Elasticity. London: 
Cambridge University Press, 1931; New York: The Macmillan Company, 

Horger, O. J., Jour, of Applied Physics, 9, 457 (1938). This article contains 
a good bibliography on the subject, 

Chap. IX] 



However, in spite of this, the experimental method of study- 
ing the stresses in many structures is easier than the theo- 
retical method, and the experimental method has the advan- 
tage over the theoretical method that it carries with it the 
conviction of a more direct appeal to nature for the informa- 
tion desired. 

Quarter-, half-, and full-wave plates. Quarter-, half-, and 
full-wave plates are made of quartz, selenite, or mica cut or 
split parallel to the optical axis. The thickness of the plate 
is made such that the relative retardation of the ordinary 
and extraordinary ray is \, J, or 1 full wave length. The 
thickness r required for a quarter-wave plate is 


4(n e - n ) 

where n e is the index of the crystal for the extraordinary ray 
and n for the ordinary, and X is the wave length in question. 
For mica the thickness of a quarter-wave plate for the D 



Elastic Limit 
(lbs. /square inch) 

Coefficient of 

Forced Double 

Refraction (Brewster's) 












Carleton, R. B., Rev. Sci. Instruments, 5, 30 (1934). 

Solakian, Arshag G., Mech. Eng., December, page 767 (1935). 

lines is about 0.036 mm. Although for mica the quantity 
(n e —n ) varies from specimen to specimen, 59 it can be taken 
as essentially constant for all wave lengths. Therefore, 
the thickness of a quarter-wave plate is roughly proportional 
to the wave length for which it is intended. 

69 Einsporn, E., Phys. Zeits.. 87, 83 (1936). 

388 OPTICS [Chap. IX 

A quarter-wave plate, when it is set perpendicular to a 
beam of polarized light with its principal directions at 45° to 
the azimuth of polarization, retards one half of the polarized 
light until its phase is 90° behind the phase of the other half, 
thus producing circular polarized light. Conversely, a 
quarter-wave plate will change circular to plane-polarized 
light. A half-wave plate, similarly oriented, transforms 
plane-polarized light to plane-polarized light rotated in 
azimuth by 90°. 

The principal directions of mica are determined by inter- 
posing it between crossed Nicols. The principal directions 
are parallel and at right angles to the azimuth of polarization 
of the incident light when the mica (of any thickness) is so 
oriented that it does not affect the cutoff of the second Nicol. 

Tutton's test 60 for distinguishing between the two principal 
directions in a quarter-wave plate is to place the plate be- 
tween crossed Nicols (with its plane perpendicular to the 
axis of the beam of incident white polarized light) oriented in 
an azimuth such that the restored light is a maximum. The 
principal directions in the plate now make angles of 45° with 
the azimuth of vibration of the incident polarized light. The 
mica plate is rotated first about one principal direction and 
then about the other, so that, in each case, light traverses a 
thicker layer of mica. In one case the color passes from 
bluish gray through iron gray to black, and in the other case 
the color passes from white to yellow and then through colors 
of a higher order. The latter color sequence corresponds to 
rotation about the principal direction of slower vibration in 
the mica and the first case corresponds to the principal direc- 
tion of faster vibration in the mica. 

Splitting of mica. Quarter-wave plates are most easily 
made from mica, since it is easily split to the thickness 

The stock sheets are split from clear mica plates. 61 The 

60 Kaplan, Joseph, J.O. S. A., 14, 186 (1927). 

61 Mica is obtainable from Eugene Munsell, 200 Varick Street, New York 

Chap. IX] 



starting sheet is trimmed to about 3 inches square with sharp 
tin snips so as to have clean edges. (The exact size of the 
starting sheet is immaterial.) One corner of the starting 
sheet is then frayed out by rubbing it, and a clean dissecting 
needle is introduced to divide the sheet approximately in 
half. A drop of water is introduced in the cavity so pro- 
duced. 62 The mica is then split all around the edges by 
working the needle along, point first, at an angle of about 30°, 
so that the first cleavage starts inside the boundary of the 
sheet. This avoids a terraced cleavage. After the needle 
has gone around the circumference, a second drop of water 
is introduced, and the plates are drawn apart. The water so 
facilitates cleavage that the sheets may be separated almost 
as easily as the pages of a book. This process is repeated 
until the thickness is approximately 0.036 mm or as thin as 
desired. Each time, the sheet is divided so as to give two 
sheets of approximately the same thickness. 

Mica gauges. 63 A gauge may be made up as shown in 
Fig. 42. To make such a gauge the principal directions are 

Strips are cut from a The strips are mounted A cover glass is 
very^hin mica on a glass plate cemented over the 

^to form steps. ^^ mica and 

^s^^V suitable 

Jabels are 


The strips «nre 
cemented in place 
with balsam. 

Fig. 42. 


first marked on a starting plate. The thinnest possible sheet 
is then split from the starting plate and cut up into strips 
about | inch wide. The strips are cut at an angle of 45° 
with the principal directions. These strips are then cut to 

62 Strong, J., Rev. Sci. Instruments, 6, 243 (1935). 

63 Wright, Lewis, Light, page 289. New York: The Macmillan Company, 



[Chap. IX 

hole for 

give rectangles with lengths of 2 inches, 1-g- inches, If inches, 
If inches, and so forth. (See Fig. 42.) The strips are next 
cemented (with balsam) between glass plates as illustrated, 
care being exercised to see that none of the strips are mounted 
upside down or rotated end for end. The steps so formed 
are then indexed. 

The retardation per step of the gauge is determined as 
follows : After the analyzer is set for maximum transmission 

of the light, the gauge is 
placed on the mirror of the 
Norremberg doubler (see Fig. 
43) either parallel or perpen- 
dicular to the azimuth of po- 
larization. A sodium light 
should be used for illumina- 
tion. The index number of 
the step which gives opacity 
is noted. The step giving 
opacity is a quarter- wave 
plate for the D lines. Other 
steps are proportionately 
greater and less. 

Using the gauge. The gauge 
is used as follows : First, the 
analyzing Nicol of the Nor- 
remberg doubler is set for ex- 
tinction. The mica of unknown thickness is placed on the 
bottom mirror of the doubler, with its principal direction mak- 
ing an angle of 45° with the azimuth of polarization to give 
maximum transmission. Then the gauge strip is laid on top 
of the mica so that it is either parallel or perpendicular to the 
azimuth of polarization. At one of these orientations, the 
steps show "interference" colors, and at the other, and proper 
one, opacity is obtained for one or two of the steps. The 
calibration value of the step which gives opacity corresponds 
to the retardation of the mica sample. Interpolation may 
be required to make a delicate measurement. 


tested -" 

Fig. 43. 

Chap. IX] OPTICS 391 

Magnification of lenses. The transverse magnification of 
a lens is the ratio of image diameter to object diameter, or, 
expressed another way, it is the ratio of transverse image 
displacement to transverse object displacement. For a 
simple lens the magnification is given by the ratio of image 
distance to object distance. For a system such as a spectrom- 
eter, which has a collimating element (lens or mirror) with 
the object at or near its focal plane and a telescope element 
also with the image at or near its focal plane, the magnifica- 
tion produced is the ratio of the focal length of the telescope 
element to that of the collimating element. 

Another case, encountered in a telescope, is that in which 
parallel light is received by the objective and observed by an 
eyepiece adjusted so that its focal plane is very near the focus 
of the objective. Here, the angular magnifying power is the 
ratio of the focal length of the objective to that of the eye- 

The longitudinal magnification of an image-forming sys- 
tem gives the ratio of the displacement of the image along 
the optical axis to the displacement of the object. In the 
case of a system composed of two lenses (or mirrors) with 
the object and image at or near the respective focal planes of 
these elements, the longitudinal magnification is given by 
the square of the ratio of the focal lengths. 

Other properties of lenses. When a beam of parallel light 
is focused with a thin lens on the optical axis, its focal length 
/ is given by the expression 



where n and r 2 are the respective radii of curvature of the 
two surfaces of the lens, and n is the index of refraction of 
the material from which the lens is constructed. The r's are 
taken positive if the curvature acts to converge the light. 

If the light is inclined to the optical axis of the lens, it 
exhibits astigmatism as shown in Fig. 44. For example, the 
best focus of a distant star, which would be a small hard spot 



[Chap. IX 

of light on the optical axis, is a soft image when the lens is 
inclined. The diameter of the smallest image is known as 
the "circle of least confusion ." Within the focal distance 
giving the smallest off-axis image, the lens gives at one 
particular distance a rather sharp line focus, which is per- 
pendicular to the plane passing through the image and the 
optical axis. Also, outside this image another rather sharp 
line focus is obtained. This line focus is perpendicular to 
the first line and parallel to the plane referred to above. 

simple lens 

On this surface 
Ihe images of 
points look like 

On this surface 
the images of 
points look like 


On this surface 
the images of 
points look like 

Fig. 44. 

The astigmatism of a simple lens is illustrated in Fig. 44. 
The locus of the inner astigmatic images is a circle, a, having 
a diameter . 


d = 

3 + U- 1 

or 0.275 / for n = 1.5, and the locus of the outer astigmatic 
images is a circle, b, of diameter 


d = = _/ . , 
1 + ft -1 

or 0.6 /for n = 1.5. 

Properties of mirrors. The mirrors generally used in 
optics are conic sections of revolution and the flat. They 
are paraboloidal for focusing parallel light, ellipsoidal for two 

Chap. IX] 



conjugate real focii, and hyperboloidal for two conjugate 
focii, one of which is virtual. The spherical mirror is, of 
course, suited for focusing light from a source at its center of 
curvature exactly back on 
the center. 

When a spherical mirror 
of radius R is used to focus 
parallel light striking it at 
an angle, the image ex- 
hibits astigmatism, and the 
lines corresponding to the 
two circles shown in Fig. 44, 
determined by the positions 
of the astigmatic images, 
are a circle of diameter R 
and a straight line, respec- 
tively. (See Fig. 21, Chap- 
ter XL) 

Properties of prisms. Some interesting properties of a 
right-angle prism are illustrated in Fig. 45. 

.This prism, viewed through the long face and perpen- 
dicular to the vertex of the 

Fig. 45. 



the object 
' to the 
of its 
angle of 
but the 
image is 

90° dihedral angle in one azi- 
muth, has the interesting and 
often useful property of re- 
turning a beam of light back 
on its path, regardless of the 
angle of incidence on the 
long face in the other azi- 
muth. Fig. 46 illustrates the 
corresponding property for 
the corner of a cube. 

Optical recording systems. 
Professor Hardy has written 
an excellent article on recording systems as applied to oscillo- 
graphs. 64 We can refer only to his results. He concludes 

64 Hardy, A. C, /. 0. S. A., U, 505 (1927). 


mirrors mutually 
perpendicular (corner of a cube) 

Fig. 46. 



[Chap. IX 

that a simple optical system with a single lens in front of the 
galvanometer mirror will give as much illumination on the 
recording film, on a basis of equal resolving power, as any 
other possible stigmatic system. Furthermore, he points 
out that the focal length of the simple systems should be 
chosen so that the limit to the resolving power is set by the 
photographic material rather than by interference effects. 
Although 25 lines/mm or more can be resolved by photog- 
raphy, Hardy sets an arbitrary practical limit of 0.1 mm as 
the resolving power of the photographic material. To ob- 
tain maximum illumination and at the same time to conserve 
on the use of photographic materials, the simple lens should 
be chosen to give a spot at least 0.1 mm wide. 

•film on 



I p* cylindrical lens -focusing an 
I\ image of the light source on 
Crt'Vthe tilm in the 

vertical azimuth 

condensing lens 

focusing an 

image of the 

light source 

on the galvanometer 


lens focusing 
an image of 
the slit^>n 
the film * 


Fig. 47. 

However, by using an astigmatic optical system such as 
the one shown in Fig. 47, it is easily possible to obtain nine 
times as much illumination as with the simple lens. Further- 
more, the astigmatic system has the additional advantage 
that rotation of the galvanometer mirror about a hori- 
zontal axis does not produce a vertical deflection of the 
image on the recording film. 

The calculation of the maximum velocity at which the 
recording spot can traverse the photographic emulsion and 
still yield a perceptible trace is treated in Chapter XI. 

Chap. IX] OPTICS 395 

This treatment includes the astigmatic case illustrated in 
Fig. 47. Owing to the recent developments in fast photo- 
graphic emulsions, the data given in Table VI, Chapter XI, 
for the various materials may be regarded as being dis- 
tinctly conservative. 

A bibliography of some of the best works on the subjects 
treated in this chapter is given in a footnote. 65 

. 65 Baly, E. C. C, Spectroscopy. New York: Longmans, Green and Com- 
pany, 1927. 

Forsythe, W. E., Measurement of Radiant Energy. New York: McGraw- 
Hill Book Company, 1937. 

Hardy, A. C, and Perrin, F. H., The Principles of Optics. New York: 
McGraw-Hill Book Company, 1932. 

Lecomte, J., La Spectre Infrarouge. Les Presses Universitaires de France, 

Meyer, Charles F., The Diffraction of Light , X-rays and Material Particles. 
Chicago: University of Chicago Press, 1934. 

Schaefer, C. L., and Matossi, F., Das Ultrarote Spektrum. Berlin: Julius 
Springer, 1930. 

Wood, R. W., Physical Optics, Third Edition. New York: The Macmillan 
Company, 1934. 


Photoelectric Cells and Amplifiers 


A. E. Whitford 


THE photoelectric cell has found an important place in 
the physical laboratory as a device for the measurement 
of the intensity of radiation, and as such may be classed with 
the thermocouple and the photographic plate. In common 
with the photographic plate its response varies with wave 
length, so that it does not measure energy directly, as does 
the thermocouple. The photographic plate, because of its 
ability to integrate extremely long exposures, can be used to 
measure smaller quantities of radiation than can the photo- 
electric cell. But the photographic plate has the definite 
disadvantage that its blackening is a complicated function 
of intensity and exposure time, necessitating a series of 
calibration exposures whose intensity ratios are known. 
Furthermore, plate grain, local variations in emulsion sensi- 
tivity, and nonuniform development place limitations on the 
precision obtainable in photographic photometry. Both 
the photocell and the thermocouple, when used with suitable 
precautions, give a response linear with respect to the inten- 
sity, and both are capable of giving more precise results than 
the photographic plate. In the infrared beyond about 
10,000 A, the thermocouple (or other heat-sensitive devices 
such as the bolometer) must be used. At shorter wave 
lengths, however, it is possible to measure much smaller 
amounts of radiation with the photoelectric cell than with 
the thermocouple. 



Limit of detection compared. An attempt to set an 
ultimate limit for any of the above-mentioned detectors of 
radiation must of necessity be approximate, because the 
working limit depends on various factors such as the angular 
size and shape of the source, the spectral distribution of the 
radiation it gives out, and the presence or absence of back- 
ground radiation. Because the stars offer a common basis 
for comparing the response of light-sensitive devices to weak 
sources, they are chosen as reference standards in this dis- 
cussion. The magnitudes of the stars form a logarithmic 
scale such that an intensity ratio of 100 times corresponds to 
5 magnitudes. Thus 1 magnitude represents an intensity 
ratio of -v^lOO, or 2.512. ... In general, the difference in 
magnitude between two stars whose intensities are I\ and I2 is 
given by rai — m 2 = 2.5 logio h/h. A candle at a kilometer 
has been found 1 equivalent to a star of visual magnitude 
+ 0.8, approximately the brightness of Altair. From these re- 
lations one may easily derive an equation giving the amount 
of light received by a telescope from a given star. If m is 
the visual magnitude of the star, d the diameter of the tele- 
scope in inches, and Q the amount of light expressed in 
lumens, 2 then it may be shown that 

2.5 logio Q = 7.57 -30 + 5 log™ d - m. 

As an example, we may compute the amount of light re- 
ceived by a 1-inch telescope from the polestar, 3 which has a 
visual magnitude of 2.1. Substitution in the formula gives 
1.5 X 10 -10 lumen. Of course, the stars differ among them- 
selves as to the spectral distribution of the light they give 
out. In this comparison, stars of spectral class G having 

1 Russell, H. N., Astrophys. J., 43, 129 (1916). 

2 The lumen is the unit of luminous flux. It is equal to the flux emitted in a 
unit solid angle by a uniform point source of 1 international candle. 

3 Polaris, in combination with a small laboratory telescope, provides a 
convenient order-of-magnitude standard source for testing the responses of 
photoelectric cells to weak radiation. It is a variable, with a period of about 
4 days, and an amplitude between maximum and minimum of 0.08 magnitude 
to 0.16 magnitude, depending on the spectral sensitivity of the measuring 
device. Its year-round availability and negligible diurnal motion are reasons 
for its choice over more constant stars. 



[Chap. X 

the same temperature as the sun are selected. As is well 
known, the radiation curve of the sun is roughly that of a 
black body at 6000 °K. The results for four detectors of 
radiation are given in Table I. These represent the limit 
reached in actual practice and not the theoretical limit. In 
the case of the photoelectric cell and thermocouple, for which 
the response is a "pointer reading/' the criterion for limit of 
detection is taken as the average deviation of successive 
deflections from the mean when measuring a star near the 
useful limit of the instrument. 

Limit of Detection for Stars of Solar Type 










(X 10" 14 ) 

(X 10" 9 ) 







Unaided human 
eye & . : 





cell c 






Thermocouple**. . 






° Ross, F. E., and Calvert, Mary, Atlas of the Milky Way. Chicago: Uni- 
versity of Chicago Press, 1934. 

b Russell, H. N., Astrophys. J., 4-5, 60 (1917). 

c Smith, S., Astrophys. J., 76, 486 (1932); ML Wilson Contr. No. 457. 

d Pettit, E., and Nicholson, S. B., Astrophys. J., 68, 279 (1938); Mt. Wilson 
Contr. No. 869. 

The data given in Table I are for response to a point source 
emitting "white" light, that is, light having the spectral 
quality of sunlight. For cases in which the image of the 
source cannot be made very small, the thermocouple must 
have larger receivers, and the limit is not as small as that 
given in the table by a factor of 5 to 10. The photographic 
plate also suffers when the source is an extended luminous 


area, because the light is spread over a greater area on the 
emulsion. According to Biltz, 4 at 4360 A it requires an 
energy of the order of 10 -2 erg/cm 2 at the emulsion surface 
to give a perceptible blackening. The response of the photo- 
electric cell is independent of image size and shape as long as 
the image is not larger than the cathode. This gives it 
considerable advantage over the photographic plate in rapid 
measurement of low surface brightness. 

The thermocouple, of course, measures energy independent 
of the spectral distribution of the light. The response curve 
of the human eye is given in Fig. 4, page 410. At the wave 
length of maximum visibility, 1 lumen = 1.61 X 10~ 3 watt. 
The variation of the sensitivity of various photographic 
plates with wave length is given by Mees. 5 The spectral 
response curves of various types of photoelectric cells are 
given on page 401. Assuming that the limit of detection is 
set by the smallest current that can be measured, these curves 
may be considered to give the limit of detection as a function 
of wave length. The case cited in Table I in which 4 X 10 -9 
erg/sec. was the limit probably represents a favorable in- 
stance. By choice of a suitable cell for each spectral region, 
the range from 2500 A to 9500 A can be covered with a limit 
of detection not greater than 10~ 7 erg/sec. 

Types of cells. Two general types of cells have been 
found useful in the physical laboratory. One is the photo- 
emissive cell, historically the oldest. In this type electrons 
are ejected from a metallic surface by the action of light and 

4 Biltz, M., Phys. Zeits., 34, 200 (1933). If the star image cited in Table I 
is assumed to be a round, uniformly illuminated spot 0.06 mm in diameter, 
the energy received at the plate is easily computed to be 0.25 erg/cm 2 , or 25 
times the figure given by Biltz. The difference may be ascribed to three 
factors: (1) Part of the radiation came in wave lengths to which the plate was 
insensitive. (2) The limit in stellar photography is set by lack of contrast 
between the star image and sky background. The sky brightness is so low 
that it does not affect other less sensitive detectors of radiation very seriously. 
(3) When the blackened area is very small, a higher density is required to 
make the image perceptible. Hubble, E., Astrophys. J., 76, 107 (1932); Mt. 
Wilson Contr. No. 453. 

8 Mees, C. E. K, J. O. S. A ., 21, 753 (1931), 22, 204 (1932), 23, 229 (1933), 
25, 80 (1935). 


are collected on an electrode maintained at a positive poten- 
tial by an external battery. The other type is the photo- 
voltaic cell, a comparatively recent development. In these 
cells light causes a transfer of electrons across the rectifying 
boundary between two dissimilar materials, such as copper 
and copper oxide, or selenium and another metal. The 
current is sent through the external circuit entirely by the 
voltage generated within the cell, and no battery is required. 

A third type of light-sensitive device, the photoconductive 
cell, depends for its action on the change in resistance of 
certain materials, such as selenium, when exposed to light. 
Though greatly improved in recent years, these cells have 
not found much application in the physical laboratory. A 
serious drawback is the nonlinear response to light. 6 

Characteristics of emissive-type cells. The spectral 
sensitivity of various types of photoemissive cells is shown in 
Fig. 1. The relative height of the various curves is only 
approximately correct, since there is considerable individual 
variation in cells of the same type. The vertical scale is 
intended to represent the average emission of good-quality 
vacuum cells obtained from commercial manufacturers. 

The cesium oxide cells are sensitive to the greatest range 
of wave lengths and are therefore probably the type most 
generally useful in the laboratory. This type of sensitiza- 
tion was developed to meet the need for a cell that would 
have a high response to light from incandescent tungsten 
bulbs, in which most of the energy comes in the red and 
infrared. Cesium oxide cells are very widely used in com- 
mercial applications of the photoelectric cell, such as for the 
reproduction of sound in motion pictures. Indeed, since 
their introduction in about 1930, the production of emission 
cells of other types has become almost negligible. 

The cesium oxide cell has one drawback when it is used for 
measuring the light from very faint sources. It has rela- 
tively large dark current, due at least in part to thermionic 

6 For further information see Henney, K., Electron Tubes in Industry. 
Second Edition. New York: McGraw-Hill Book Company, 1937. 

Chap. X] 



emission from the sensitive surface at room temperature. 
Currents as large as 10~ 9 ampere have been reported. 7 
Present-day cells, when of a design which minimizes insula- 
tion leakage over the bulb and base, usually have a dark 
current of the order of 10 -11 to 10 -12 ampere. If the photo- 
current is very much (say 1000 times) smaller than this, it 

3000 4000 5000 6000 7000 8000 9000 10,000 
wave length in Angstrom units 

Fig. 1. Spectral sensitivity of various types of photoelectric cells. 

will be masked by unavoidable irregularities in the dark cur- 
rent. The dark current can be reduced to 10~ 15 ampere or 
less by refrigerating the cell with solid carbon dioxide. This 
has been done by Hall 8 and Bennett 9 in their application of 
cesium oxide cells to the photometry of stars. Details of the 
design of such an arrangement are given on page 424. 
In certain applications, the high infrared sensitivity of the 

7 Kingsbury, E. F., and StiUweU, G. R., Phys. Rev., 37, 1549 (1931). 

8 Hall, J. S., Astrophys. J., 79, 145 (1934). 

9 Bennett, A. L., Pub. Am. Astr. Soc, 8, 209 (1935). 


cesium oxide cell is inconvenient. For example, in compar- 
ing the color of certain objects, it may be desired to measure 
the intensity through a blue filter. Almost all blue filters 
made of glass or gelatin are more or less transparent to the 
infrared, so that with a cesium oxide cell large error will be 
made unless an additional filter to remove the infrared is 
used. The standard filter for removing the infrared is a 
solution of copper salt. 10 Two special glass filters are also 
satisfactory. These are the Corning Glass Company's Aklo 
and the Jena BG-18. 11 

The potassium oxide cell is valuable in applications to 
colorimetry and photometry when infrared sensitivity is 
undesirable. Ordinary filters may be used with it, without 
particular attention to infrared leaks. It has a fair degree 
of red sensitivity, with a threshold at about 8000 A, and a 
higher green and blue sensitivity than the cesium oxide cell. 
The thermionic emission is negligible. Cesium-magnesium 
cells are also useful for applications of this type. The thresh- 
old for this type of surface is at about 7000 A. 

Potassium hydride cells are sensitive to a fairly narrow 
range of wave lengths, mainly in the blue, with a maximum 
at about 4400 A. Their sensitivity to white light and par- 
ticularly to light from incandescent tungsten is considerably 
lower than for cesium oxide cells. However, in the pho- 
tometry of stars, which are in general much hotter than in- 
candescent lights and give out correspondingly greater 
amounts of blue light, potassium hydride cells have been 
found extremely valuable. For this work, an important 
advantage is their extremely low dark current, limited, it 
seems, only by the insulation of the bulb. Smith 12 has 
reported a dark current of only 5 X 10~ 18 ampere for a 
potassium hydride cell in a f used-quartz bulb. 

10 Gibson, K. S., J.O.S.A., 13, 267 (1926), recommends 57 g of CuS(V5H 2 
to 1 liter of water; about 2 cm is required. A transmission curve is given in 
Fig. 24, Chapter IX. 

11 Jena glass niters are obtainable from the Fish-Schurman Corporation. 
250 East 43rd Street, New York City. 

12 Smith, S., Astrophys. J., 76, 486 (1932); ML Wilson Contr. No. 457. 


For the ultraviolet the cesium oxide cell can be used. For 
applications in which the cell must be sensitive only to the 
ultraviolet, several types of cells are available, with different 
thresholds. Sodium cells are sensitive mainly to wave 
lengths in the range 2000 A to 4000 A, with a slight response 
to visible light. For the shorter wave-length portion of the 
ultraviolet a thorium, titanium, or tungsten cathode may be 
used. Thorium is sensitive from 2500 A to 3600 A, titanium 
from 2500 A to 3200 A, and tungsten from 1700 A to 2700 A. 
Insufficient data are available for plotting spectral response 
curves of these tubes along with those given in Fig. 1, but 
the response is believed to be comparable to that of a sodium 
cell at wave lengths below 3000 A. These cells are especially 
useful in investigations of the biological effects of radiation. 

In Table II is collected a representative list of photo- 
electric cells of the emission type obtainable from American 
manufacturers. When several cells which differ only as 
regards size of cathode or type of base are available, only 
one is listed. The sensitivity rating (except as otherwise 
noted) is based on the response to light from incandescent 
tungsten at a color temperature of 2870 °K., which has been 
tentatively adopted as a standard source for comparing 
photocells. This is roughly the operating temperature of a 
300-watt gas-filled tungsten bulb designed for general illumi- 
nation purposes. When tested with a bulb at a lower 
temperature, cesium oxide cells will give a slightly higher 
apparent rating, but the rating of the blue-sensitive cells 
will be affected unfavorably. For rough tests gas-filled 
Mazda bulbs of from 50 watts to 100 watts may be used. 
At normal voltage these lamps have an efficiency of approxi- 
mately 1.0 candle power/watt (within 10 to 20 per cent) and 
operate at about 2700°K. 13 

The insulation between cathode and anode is an important 
factor when a cell is to be used to measure a very faint source. 
If the insulation is not good enough, the dark current due to 

13 For further information see Moon, P. H., Scientific Basis of Illuminating 
Engineering. New York: McGraw-Hill Book Company, 1936. 



[Chap. X 

Photoelectric Tubes 


Type Number 



Type of 



on Bulb 
(if any) 

PJ-14 a 





































71-TA 6 







1038-A C 























2 seals 





















920 d 












CE-15 6 










































0.10 e 







0.02 e 







0.001 e 

















Typical sensitivity is rating when exposed to tungsten light at 2870 °K., 
except as otherwise noted. Std base means regular four-pin radio base. 
When no terminal on bulb is mentioned, both terminals come out through pins 
in the base. Manufacturers: GE, General Electric Company, Schenectady, 
New York; GM, G-M Laboratories, 17S1 BeJmont Avenue, Chicago, Illinois; 
RCA, RCA Manufacturing Company, Harrison, New Jersey; WH, Westing- 


house Lamp Company, Bloomfield, New Jersey; CE, Continental Electric 
Company, Geneva, Illinois. 

a Listed as "special high sensitivity cell similar to PJ-14." 

6 High insulation cell with internal guard ring. 

c Quartz bulb; terminals through common seal, but special sheath provided. 

d Twin cell. Two cathodes and two anodes in single bulb. 

e Response to total radiation from S-l lamp in Corex bulb, with standard 
reflector, at a distance of 1 foot. 

f Quartz bulb; also available in gas-filled type. 

Not manufacturer's figure; estimated from other data. 

h As far as is known, this type of cell is not now regularly produced by any 
American commercial manufacturer, although it might be obtained on a 
special order. Those used in astronomical photometry have been made by 
Professor J. Kunz, Department of Physics, University of Illinois. The 
General Electric Company, Ltd., Magnet House, Kingsway, London, W. C. 2, 
manufactures a gas-filled potassium cell known as the Osram KG-7. 

leakage may be many times larger than the photocurrent. 
For example, in certain cells designed for use in the sound 
head of motion-picture projectors, in which the photocurrent 
may be on the order of 1 microampere, the two terminals 
come out to pins on a standard four-prong radio-tube base. 
In general, cells of this type are not satisfactory for currents 
smaller than about 10~ 10 ampere. In some cells the connec- 
tion to either the cathode or the anode is made via a separate 
cap on the bulb, and thus a considerable length of clear glass 
bulb is interposed between the terminals to serve as insula- 
tion. Cells without a base in which the terminals come out 
through a common pressed seal are better than those with a 
base, but not as good as those in which there are two seals 
at opposite ends of the bulb. 

In many cases the leakage current over the surface of the 
bulb can be practically eliminated by a guard ring at a proper 
place on the cell. This can be made by wrapping a few 
turns of fine wire around the cell and painting with aquadag 14 
or a mixture of lampblack and mucilage. The guard ring 
should be connected to a point in the circuit such that the 
potential difference across the insulation on the cell will be as 
near zero as possible. In some processes of manufacture, 
particularly of cesium oxide cells, a thin deposit is left on the 

14 Aquadag is obtainable from the Acheson Colloids Corporation, Port 
Huron, Michigan, 


inside walls of the bulb, which greatly reduces the insulation 
resistance. The obvious remedy is an internal guard ring, 
and some manufacturers regularly provide this feature on 
certain of their cells. 

In general, insulation leakage currents are greatly reduced 
by cooling, because of the quasi-electrolytic nature of such 
conduction. For instance, the dark current of an RCA 917 
cell fell from about 10 -10 ampere to about 10 -13 ampere when 
cooled with dry ice. The latter current is not a serious 
detriment except when the highest sensitivity is required. It 
thus becomes possible to use an inexpensive commercial cell 
to measure rather faint sources. A suitable arrangement 
for refrigerating cells with dry ice is described on page 424. 

The manufacture of cells. The prospective user of photo- 
electric cells will find it by far the most economical procedure 
as regards both time and money to purchase the cells from 
an established manufacturer. The production of highly 
sensitive cathodes is still very much an art which has never 
been fully described in the literature and requires some 
experience to master. Most manufacturers will accept 
orders for special cells with internal guard rings or other 
modifications required in particular applications. 15 

Vacuum and gas-filled cells. In vacuum cells the anode 
merely collects all of the electrons ejected from the light- 
sensitive cathode. The current for a given light increases 
rapidly with voltage up to about 25 volts and then gradually 
becomes saturated and does not increase further. The 
introduction of an inert gas at a pressure of a few tenths of a 

15 For workers who wish to experiment with the making of cells, the follow- 
ing references will give fairly complete instructions: 

Hughes, A. L., and DuBridge, L. A., Photoelectric Phenomena. McGraw- 
Hill Book Company, 1932. Chapter 12 gives a summary of methods for all 
types of cells. 

Nottingham, W. B., Frank. Inst, J., 206, 637 (1928). Details for alkali- 
hydride cells. 

Prescott, C. H., Jr., and Kelley, M. J., Bell System Techn. J., 11, 334 (1932). 
Detailed analysis of the process for cesium oxide cells. 

Rentschler, H. C, Henry, D. E., and Smith, K. O., Rev. Set. Instruments, 3, 
794 (1932). Deposition of thorium, tungsten, and many other metals on the 
cathode by a sputtering process. 


millimeter permits an amplification of the original photo- 
current due to ionization by collision. Fig. 2 shows the 
relation between the voltage and the current for the same 
sensitive surface in a vacuum and with gas. With increasing 
voltage the gaseous amplification factor increases until at a 
certain voltage a self-maintaining glow discharge sets in, 
which, if continued for more than a few seconds, may seri- 
ously damage the sensitive surface, The glow voltage is 
lower the greater the illumination. Thus a cell which is 


1 so 



+- 4- 

/ + 




p 40 





f - 

-"^"^ + 

+ vacuum + 


O 20 40 60 SO 10O 120 


Fig. 2. Volt-ampere characteristics of vacuum and gas-filled photoelectric 


safely below the glow voltage in the dark or for low illumi- 
nation levels will glow when exposed to strong light. A 
gaseous amplification factor of 10 is about the maximum 
usually recommended. The gas pressure in commercial cells 
is usually adjusted so that 90 volts is the maximum safe 
voltage on the cell for illumination of the order of 0.1 lumen. 
However, when used with faint sources, this may be exceeded 
somewhat and an amplification factor of 20 to 40 realized. 
When a voltage increase of 10 volts doubles the response, the 
safe limit has been reached. Amplification factors as high 
as 600 have been reported, 16 but a gas cell becomes very 
unstable when too near the glow voltage. 

16 Steinke, E., Zeits.f. Physik, 38, 378 (1926). 


A protective resistance of at least 100,000 ohms 
should always be in the circuit of a gas-filled cell to limit 
the current to nondestructive values in case of a glow 

The speed of response of a vacuum cell is limited only by 
the transit time of the photoelectrons, though usually the 
amplifier sets the limit. Gas-filled cells, however, have a 
definite time lag. When used with a modulated light signal, 
the response falls off gradually at the higher audio fre- 
quencies, the reduction becoming serious at 10,000 cycles. 
Data on the frequency response characteristic are supplied 
by cell manufacturers, particularly for cells to be used in 
sound reproduction. 

When operated on the steep part of the current- volt age 
curve, gas cells may give a nonlinear response to light. The 
voltage across the cell is reduced when exposed to light by 
the amount of the potential drop in the external resistance, 
and with reduced voltage, the cell is less sensitive. On the 
other hand, the gaseous amplification factor is greater for 
more intense illumination, causing an error in the opposite 
sense. No general rule can be laid down; each situation 
must be analyzed separately. In most laboratory applica- 
tions, however, in which the intensity is usually low, diffi- 
culties with nonlinear response are less likely to be encoun- 
tered. In fact, most of the drawbacks of gas-filled cells 
become less serious when used with faint sources. Recently 
Stebbins and Whitford 17 calibrated a gas-filled potassium 
hydride cell over a thousandfold range in intensity at a con- 
stant cell voltage. The largest current was about 3 X 10 -12 
ampere. They found no departures from linearity signifi- 
cantly greater than the probable error, which was about 
1 per cent. 

Photoelectric currents are feeble enough at best, and the 
gain provided by gaseous amplification is often just the 
margin of safety between satisfactory and unsatisfactory 

17 Stebbins, J., and Whitford, A. E., Astrophys. J., 87, 237, 1937; Mt. Wilson 
Contr. No. 586. 

Chap. X] 



operation. However, when the light intensity is great 
enough, vacuum cells are to be preferred because of their 
greater stability. 

Photovoltaic cells. In its usual form the photovoltaic cell 
consists of a thin metallic disk, coated with a film of sensitive 
material, sealed in a moisture-proof case with a glass window, 
and provided with suitable terminals. On some cells the 
terminals are two pins spaced to fit two of the holes of a 
standard four-prong radio socket. Cells of the photovoltaic 
type are manufactured by the General Electric Company, 
G-M Laboratories, Inc., Westinghouse Electric and Manu- 
facturing Company (Photox), and Weston Electrical Instru- 
ment Corporation (Photronic cell). 

Fig. 3 shows the relation between current output and 
illumination for the Weston Photronic cell with various 

Response of Photronic cell as a function of illumination for 
values of the external resistance. 

external resistances. The total current generated by the 
light is believed to be proportional to the light intensity. 
However, this current divides between the internal and 
external resistance. The internal resistance of the cell 
is about 7000 ohms in the dark, decreasing rapidfy with 



[Chap. X 

increasing iflumination. 18 As the figure shows, the result is 
a nonlinear current output unless the external resistance is 
quite small. With a low-resistance meter, however, the out- 
put is practically linear and amounts to 120 microamperes/ 
lumen for a tungsten lamp at 3000 °K. 

The spectral sensitivity curve of the Photronic cell is 
shown in Fig. 4, along with the sensitivity curve of the 

3000 4000 5000 600O 

wave length in Ang-strdm units 



blue green orange red 



Fig. 4. Spectral sensitivity of Photronic cell, compared with that of the 

human eye. 

human eye. By use of a suitable filter, the response curve 
of the cell may be modified to match that of the eye very 
closely. Such a filter is supplied by the makers of the cell. 
The photovoltaic cell, when used with a sensitive d'Arson- 
val galvanometer, gives very stable and reproducible deflec- 
tions. If the Photronic cell mentioned above were used 
with a galvanometer having a current sensitivity of 2 X 10~ 10 
ampere/mm, 1 mm would correspond to 1.7 X 10~ 6 lumen. 
The sensitive area of the cell is about 0.012 square foot. 
From this it may be computed that 1 mm deflection would 

18 Romain, B. P., Rev. Set. Instruments, 4, 83 (1933). 


correspond to the amount of light received by the cell from a 
standard candle at 85 feet. A slight gain might be realized 
by using a more sensitive galvanometer, but it is not worth 
while to push the sensitivity to the extreme limit, because of 
the comparative ease with which much smaller amounts of 
light may be measured with an emission-type cell and 

Amplification of the output of photovoltaic cells is not 
feasible, because of their low voltage sensitivity. 19 Because 
of the high capacity (0.5 microfarad) between the termi- 
nals, the response to modulated light intensities falls 
rapidly with increasing frequency, and the cell is not well 
adapted to sound reproduction. The power sensitivity is 
high and a sensitive relay can be operated directly on the 
output if there is a change in illumination of 0.2 lumen or 
more. 20 In many applications of the photoelectric cell to 
automatic control mechanisms, in which only an on-and-off 
signal is required, this is a simple and convenient arrange- 

For laboratory measurements, the photovoltaic cell is 
recommended in applications in which there is sufficient 
intensity available, because of its simplicity and compact- 
ness, and because it does not require an external battery. 

Amplification of photoelectric currents. Photoelectric 
cells of the emission type may be used with galvanometers 
down to about the same limit of light intensity as that given 
above for photovoltaic cells. For fainter sources some more 
sensitive current-measuring device must be used. Electrom- 
eters of various types can be used, and in certain applica- 
tions may be the most desirable instrument. In particular, 
the Lindemann electrometer, or the Cenco-Dershem modifi- 
cation of it, is useful when light weight and compactness are 
important considerations. In the last few years, however, it 
has increasingly been the practice to amplify small photo- 

19 Wilson, E. D., Rev. Sci. Instruments, 2, 797 (1931). 

20 Suitable relays are manufactured by the Weston Electrical Instrument 
Corporation and the G-M Laboratories, Inc. 



[Chap. X 

electric currents up to the level at which they can be read on 
a galvanometer. The advantages of amplification are that 
(1) currents which must be measured by the rate-of-drift 
method with an electrometer can be measured by the more 
convenient steady-deflection method, and (2) the amplifier 
is more rugged and portable than the electrometer. But 
for attaining the ultimate limit in measuring photoelectric 
currents, the Hoffmann electrometer is probably still the 
best instrument. 

In case the photoelectric current is varying rapidly with 
time, the variations can easily be amplified many thousand- 
fold by means of a multistage a.c. amplifier, the technique 
for which has been highly developed because of numerous 
applications in the radio, telephone, and motion-picture 
industries. In the laboratory, however, it is usually desired 
to measure the photoelectric current due to a steady source. 
Hence d.c. amplification must be used. 

Direct-current amplifiers. The fundamental circuit of a 
single-tube d.c. amplifier is shown in Fig. 5. The photo- 
current passes through a resistance R Q , and the resulting 



Uf|i|i|ip— h|i 

Fig. 5. Direct-current amplifier for photocurrents. 

voltage drop alters the grid potential of the tube. The 
consequent change in plate current is read on the galva- 
nometer. The normal plate current of the tube is balanced 
out by adjustment of R x so that the galvanometer reads zero 
when there is no light on the cell. Since R x is usually at 
least fifty times the galvanometer resistance, the galva- 

Chap. X] 



nometer will indicate very nearly the entire change in plate 
current. If % is the photocurrent, the galvanometer current 
%i is then %i = ioRog m , where g m is the mutual conductance of 
the tube. The mutual conductance, more correctly called 
the grid-plate transconductance, is defined by the relation 
9m = di p /de g , where i v is the plate current and e g is the grid 
voltage. To obtain high current amplification it is necessary 
to make R as great as possible. It is useless, however, to 
increase R indefinitely, because it is shunted by the grid-to- 
filament resistance of the tube itself. With most tubes 
designed for use in radio receivers this resistance is not over 
10 8 ohms. The instability of the plate current is such that 
it is not worth while to use a galvanometer with a sensitivity 
better than 10 -8 ampere/mm. Such a circuit is therefore 
limited to measuring currents greater than 10 -12 ampere. 

The comparatively low value of the grid-to-filament 
resistance in ordinary tubes is due not so much to poor 


Recommended Operating Conditions and Other Essential Data for 
Two Makes of Electrometer Tubes 

Filament voltage 

Filament current 

Space-charge grid voltage 

Plate voltage 

Control grid voltage 

Input resistance 

Control grid current 

Plate current 

Plate resistance 

Mutual conductance 


2.5 volts 

0.09 ampere 

4 volts 

6 volts 

4 volts 
10 16 ohms 
10 -15 ampere 
60 microamperes 
45,000 ohms 
20 microamperes/volt 


1 volt 

0.27 ampere 

4 volts 

4 volts 

3 volts 
10 16 ohms 
10~ 15 ampere 
85 microamperes 
25,000 ohms 
40 microamperes/volt 

insulation as to charges reaching the grid inside the tube. 
Any variation of the grid current with voltage constitutes a 
conductance. Metcalf and Thompson 21 made a systematic 
study of the sources of current to the grid and methods of 

21 Metcalf, G. F., and Thompson, B. J., Phys. Rev., SO. 1489 (1930). 



[Chap. X 

eliminating or reducing them. As a result, a new tube 
known as the FP-54 was developed especially for the ampli- 
fication of small direct currents. It is made by the General 
Electric Company. The Western Electric Company makes 
a similar tube, known as the D-96475. These tubes have an 
inner space-charge grid to shield the control grid from positive 
ions emitted by the filament. They are operated at a very 
low plate voltage to avoid ionization of the residual gas. 


If - 0.27 ampere 

E p - 4 volts 

+■ ■+ + *■ 
E s - 4 volts 




150 $>> W 

100 ^g 
50 a I 





O 3 
U O 

-6 -5 -4 -3 

control- grid voltage 

control- grid volta-ge 







O C 
J- fc> 
■H i. 
C i. 

o o 

% ig. 6. Plate- and grid-current characteristics of Western Electric D-96475 


Because they replace an electrometer, they are often called 
electrometer tubes. Their grid resistance is approximately 
10 16 ohms. 

The recommended operating conditions and other essential 
data for the two makes of electrometer tubes are shown in 
Table III. 

Characteristic curves showing the plate current and con- 
trol grid current of a typical D-96475 tube are set forth in 
Fig. 6. The slope of the grid-current curves gives the grid 

Chap. X] 



conductance, and the reciprocal of the slope is the grid re- 
sistance. The control grid is operated at — 3 volts, because 
the curve is nearly flat at that point. The slope of the plate- 
current curve gives the mutual conductance. The curvature 
of the plate-current curve is quite noticeable, and is sufficient 
to cause appreciable nonlinearity if the grid voltage changes 
by 0.1 volt, a rather large change in most applications. 

photo- JL 
electric __ 
cell ~ZZ 

R.o=10 9 tol0"ohms 

R, =20,000 ohms 

R_2 = 10'ohm potentiom eter 

R.3 = 100 ohms 

It4 = 400-ohm potentiometer 

Us = 500 ohms 
R6 = 6 ohms 
E = 1.5-volt 



All tube potentials are supplied from two 
6-volt lead storage batteries 

Fig. 7. Circuit for amplification of photocurrents using the D-96475 tube. 

Fig. 7 gives in detail the constants of a simple circuit 
employing the D-96475 for the measurement of the photo- 
electric currents. By choosing a resistance R of the proper 
value, and varying the galvanometer shunt, a very wide 
range of currents may be covered. The calibration circuit 
in the control-grid lead provides a means of testing the 
sensitivity, which should be of the order of 100,000 mm/ 
volt for a galvanometer sensitivity of 4 X 10 -10 ampere/mm. 

The stability of the plate current in such a circuit of course 
determines the smallest current that may be measured. A 
slow drift in the galvanometer zero will not cause a serious 


loss in accuracy if the rate is constant during the time of one 
observation, since it may easily be averaged out by taking 
alternate zeros according to a definite schedule. In certain 
applications of electrometer tubes, such as measurement of 
a particles, when continuous registration over long periods is 
required, a very low drift rate is necessary. In any case it is 
desirable to reduce drift as much as possible. 

Random fluctuations of the zero, on the other hand, place 
a definite limitation on the useful sensitivity of the amplifier. 
These may be due to external conditions, such as mechanical 
vibration, poor insulation, poor contacts, stray electric and 
magnetic fields, or residual atmospheric ions settling on the 
grid lead. By suitable precautions, disturbances due to 
such causes can be practically eliminated. There remain 
inherent and unavoidable fluctuations caused by variations 
in filament emission and by shot effect and Johnson effect 
(thermal noise) in the grid circuit. In actual practice, un- 
even filament emission is found to be unimportant compared 
to the two latter sources of fluctuations. A quantitative 
discussion of these factors is reserved for a later paragraph. 

Contrary to statements which have appeared in the litera- 
ture, battery potential fluctuations are not a limiting factor 
in using a simple circuit of the type shown in Fig. 7. The 
stability is determined in the grid circuit, as is shown by the 
fact that the zero is much steadier with the high resistance 
Ro shorted out. The effects of variable filament emission 
remain. To gauge the effect of the battery fluctuations only, 
the tube may be removed and replaced by fixed resistances 
equivalent to the filament resistance and to the static fila- 
ment-to-plate resistance. The galvanometer will then be 
perfectly steady, except possibly for a slow uniform drift, 
showing that there are no sudden changes in battery voltage 
large enough to affect the stability. 

One cause of drifts is temperature variation, which changes 
battery voltages and the resistance of various parts of the 
circuit. If the temperature is constant, the galvanometer in 
the circuit shown in Fig. 7 drifts slowly in the direction of 

Chap. X] 



decreasing plate current. This is caused mainly by a de- 
crease in the filament current as the battery discharges. The 
larger the filament battery the slower the drift — hence the 
paralleling of cells in Fig. 7. As soon as equilibrium is 
reached, the rate of drift is less than 1 mm/minute at a 
sensitivity of 100,000 mm/volt. For most laboratory meas- 
urements such a rate is not a serious drawback. 

The simplest method of controlling drift is to introduce a 
counterdrift whose rate can be adjusted. Hafstad 22 placed 

photo- — 
electric — 
cell ; 

The value of the bleeder resistor R 7 depends 
on the capacity of the cell E\ 

Fig. 8. 

Control of drift by introducing adjustable counterdrift in plate 
circuit. Resistances have the same values as in Fig. 7. 

a countercell in the filament circuit and bled it with a vari- 
able resistance. A smaller cell is required if the counter- 
drift is introduced in the plate circuit. An arrangement of 
this type, similar to one described by Bearden, 23 is shown in 
Fig. 8. Part of the voltage which balances out the plate 
current of the tube comes from the cell E 1} which is being 
discharged through R 7 . The discharge rate is set so that if 
all the voltage for balancing out were taken from E h the 
drift would be in the direction of increasing plate current. 
Then by varying the amount of voltage taken from the 
loaded and unloaded cell, the normal drift due to decreasing 

32 Hafstad, L. R., Phys. Rev., U, 201 (1933). 

23 Bearden, J. A., Rev. Sci. Instruments, 4, 271 (1936). 



[Chap. X 

plate current may be compensated. The advantage of this 
method is that the drift rate may be changed immediately 
without altering anything in the circuit that requires some 
time to reach a new equilibrium. The cell E\ may be a 
flashlight cell, or if a rechargeable cell is desired, an Edison 
storage cell. In cases where the tube is to be operated con- 
tinuously over long periods, the cell Ei may be eliminated 
and the battery discharging resistor attached directly to the 
last cell of a storage battery. The drift rate is then regu- 
lated by varying the discharge current. This requires some 
patience to adjust. 

Numerous circuits have been devised with a view to 
making the plate current of a vacuum tube independent of 
small changes in supply voltage. The former two-tube 
circuits 24 required tubes with matched characteristics, which 
are often difficult to obtain. Recently practically all bal- 
anced circuits have consisted of a single tube in a suitable 
resistance network. 

The single-tube circuit originated by DuBridge and 
Brown 25 has been successfully used in many laboratories. 


Fig. 9. DuBridge and Brown's balanced circuit for the FP-54 tube. 

It is shown in Fig. 9. It can be regarded as a Wheatstone 
bridge in which R x and R 2 form two of the resistance arms, 

24 DuBridge, L. A., Phys. Rev., 87, 392 (1931). 
Wynn-Williams, C. E., Phil. Mag., 6, 324 (1928). 

25 DuBridge, L. A., and Brown, H., Rev. Set. Instruments, 4, 532, 1933. See 
Turner, L. A., Rev. Sci. Instruments, 4, 665 (1933), for other circuits. 


and the filament-to-plate resistance and filament-to-space- 
charge-grid resistance form two other arms. The resistance 
R 3 may be considered a part of the tube resistance. A 
condition that the galvanometer current be zero is obviously 

Rilp = Ril*. 

In order for this condition to remain satisfied when the bat- 
tery voltage changes, or the filament emission changes for 
any reason, it is further necessary that 

dl p _ Ri dl s 
dl f R 2 dl f 

For these conditions to be satisfied in general, it would be 
necessary for the I v versus // and I s versus 7/ curves to be 
straight lines intersecting at a common point on the // axis. 
Of course, the tubes do not have this characteristic, but over 
short ranges the tangents to the I p and I s curves do satisfy 
this condition, and it is possible to adjust the resistances in 
the circuit so that this can be made to occur at approximately 
the rated filament current. 

The resistance Rs is necessary to provide a voltage drop of 
2 volts, since the space-charge-grid voltage of the FP-54 is 
4 volts and the plate voltage is 6 volts. For the D-96475 
tube, in which both the space-charge-grid and the plate 
operate at 4 volts, it would be omitted. In a typical setup 
with the FP-54 tube, R& was 45 ohms, R z was 4000 ohms, R 2 
was 2000 ohms, and Ri was a 10,000-ohm rheostat, with R x ' 
a 50-ohm rheostat for fine adjustment. Ri was a 50-ohm 
potentiometer. The procedure in balancing the circuit is as 
follows: With the galvanometer shunted to one tenth or 
one hundredth of its full sensitivity and R\ adjusted so that 
the galvanometer reads zero when I f is near its rated value, 
If is slowly varied by means of the rheostat R 5 . With the 
galvanometer connected so that a positive deflection is 
caused by a decrease in the plate current, the deflection 
should pass through a maximum value for some value of //. 
If the galvanometer deflection goes off scale before the maxi- 


mum is reached, it may be brought back by an adjustment 
of R lt 

If the value I f for maximum is not within a few per cent of 
the rated value for the tube, the adjustment of i? 4 and Rx will 
bring the balance point to a different value //. The adjust- 
ment is finally made with the galvanometer at full sensitivity. 
Each adjustment will require a few minutes' waiting for anew 
thermal equilibrium to be established. 

The advantage of this ingenious circuit is that any change 
in the filament emission due to variation in battery voltage 
or deactivation of the filament is compensated. Also short- 
period fluctuations in the filament emission are balanced out, 
and the stability with R shorted out is somewhat improved 
over that obtained with the circuit of Fig. 7. 

Some workers have experienced difficulty in obtaining the 
balanced condition for tubes of different characteristics. 
Penick 26 has given a thorough analysis of balanced circuits, 
with special application to the D-96475 electrometer tube. 
He suggests a modification of the DuBridge and Brown 
circuit which in effect amounts to attaching the leads from 
the plate and space-charge-grid leads of the bridge to sepa- 
rate taps on the resistance E 4 . In practice this is best done 
by using two potentiometers in parallel. This introduces an 
additional element of flexibility in the circuit and enables 
a balance point to be reached for tubes of widely different 
characteristics. However, there are considerable individual 
variations in electrometer tubes, and sometimes with a 
particular tube it may not be possible to reach a balance 
point with reasonable values for the other circuit constants. 

The best type of circuit for a particular application must 
be decided upon the basis of the conditions of use. The 
limiting sensitivity obtainable is about the same for all types 
of circuits. A balanced circuit is probably preferable when 
the use is to be irregular or intermittent because of much 
shorter warm-up time. A simple uncompensated circuit 

26 Penick, D. B., Rev. Sci. Instruments, 6, 115 (1935). Full references to the 
literature are given. 


requires several hours to reach equilibrium. However, if it 
is to be used daily, there is no objection to letting the filament 
run continuously. 

In applications in which the circuit is to be used for both 
steady deflections and rate-of-drift measurements, the bal- 
anced circuit becomes rather complicated, because whenever 
the control-grid bias is changed, in order to find equilibrium 
potential the entire circuit must be rebalanced. 

In order to reach high sensitivity by the steady-deflection 
method the resistance R must be made very large. If it is 
much larger than 10 11 ohms, the time constant of the input 
circuit becomes unduly long, and considerable time is wasted 
in waiting for deflections to reach their final value. The 
highest sensitivity is best obtained by the rate-of-drift 
method. To do this the anode of the photoelectric cell (or 
other sources of the current to be measured) is connected 
directly to the control grid of the tube with no other resist- 
ance. A high-insulation switch free from contact potentials 
is necessary for grounding the grid. When the tube is 
operated at the rated control-grid voltage, the grid current is 
approximately 10~ 15 ampere. If the grid is "floated" by 
opening the grounding switch, it will draw this current and 
drift in the direction of less negative grid potential. In 
order to eliminate this drift, the grid bias must be changed to 
such a value that the positive and negative components of 
the grid current are equal and the grid current is zero. When 
operating at this so-called "equilibrium potential," the grid 
resistance, as shown by the slope of the lower curve in Fig. 6, 
is considerably reduced but is still usually as high as 10 14 
ohms and is quite satisfactory for use in the rate-of-drift 
method. The equilibrium potential is found by observing 
the drift when the grounding switch is opened for various 
values of the control-grid potential until a value sufficiently 
near zero drift is reached. The galvanometer must be 
brought back to zero after each change in the grid bias, by 
adjustment of the resistance in the plate circuit. The pro- 
cedure for measuring a small photoelectric current is then 


to observe the drift rate with the light alternately on and 
off. This may be done by timing the drift over a particular 
interval of the galvanometer scale, but better results will be 
obtained if the galvanometer deflections are continuously 
recorded photographically. The value for the slope of the 
drift curve does not then depend on observations at two 
particular points. In fact, in all measurements in which the 
inherent fluctuations set the observing limit, improved 
accuracy will be obtained with photographic registration. 26a 

Measurements made by the rate-of-drift method may be 
connected with steady-deflection measurements by measur- 
ing the same current by both methods. In effect this con- 
stitutes a measurement of the capacity of the grid circuit. 

Experimental details. The photoelectric cell and elec- 
trometer tube must be enclosed in a light-tight metal box. If 
high sensitivity is required, the container should be evacu- 
ated in order to eliminate the effect of residual ions in the air 
caused by cosmic rays. A suitable form for the container 
consists of a brass cylinder with a light window on the side. 
The amplifying tube and photoelectric cell are best supported 
entirely from one end plate in order to facilitate removal for 
adjustment. The vacuum seal may be made with a rubber 
gasket. A vacuum of the order of 1 mm of mercury is 
sufficient. The necessary switching arrangements can be 
operated by means of cams on a shaft which turns through a 
conical ground joint in the end plate. A separate metal 
shield over the tube inside of the container is advisable, and 
this may be made to serve also as a light shield to screen the 
photoelectric cell from the light of the filament. In case 
the container is not evacuated, it should be kept dry by the 
use of calcium chloride or phosphorus pentoxide. 

A satisfactory design for mounting the cell and tube in an 
evacuated brass tank is shown in Fig. 10. The diagram is 
largely self-explanatory. The construction of vacuum tanks, 
valves, and gaskets is treated in Chapter III. 

26a A suitable optical arrangement for photographic recording of galva- 
nometer deflections is shown in Fig. 47, Chapter IX. 

Chap. X] 



When a cesium oxide cell must be refrigerated to reduce 
the thermionic emission, the design shown in Fig. 11 meets 
the requirements. Here the problem is to have the cell as 
cold as possible, but to keep the light window from fogging 

Spring dick. to 
d .switch in 



for evacuating 
case and also 
operating sensi 
tivity switch 

connection for 

K | 
airhole v ^ 
to .stopcock j 

D 96475 
tube in 

studs to | 
operate . 
switch ' 

phosphor bronze 
-springs ' 
mounted oft 
Bakelite block 

resistors *~ 






^ * 

Fig. 10. Method of mounting photocell and electrometer tube in evacuated 
brass tank. (If circuit is grounded at the usual point, as in Fig. 7, the Bakelite 
insulation of the phosphor-bronze springs is unnecessary. They may be 
attached directly to the brass post as shown in Fig. 11.) 

and the wax seals from becoming too cold and brittle. This 
necessitates some rather steep thermal gradients, which are 
successfully withstood by the Pyrex tube around the cell. 
To maintain isothermal conditions, the dry-ice compartment 
is made of sheet copper, and the Pyrex tube is sheathed with 



[Chap. X 

wax-sealed leads 

to galvanometer; <■ 
power supply, 


rubber gfashet 

hard rubber 

spring contacts 
to permit removal 
of amplifier 

heatingf coil 
using 6v-5w 
to prevent 
fogging of 

** to switch handle 

picein seal 
wooden box 

2ft hair felt 

-< to stopcock 
and vacuum 

cam for operating 
switch 5 

enclosure for 


platinum switch 

| shield to prevent 
| stray light from 
entering- enclosure 


Pyrex tube 
■< picein seal 




dr^ ice' 

door for 

"dry ice" 

copper box 
and shield 
to obtain 

spring clips 
to hold 


scale of inches 

Fig. 11. Evacuated container for photocell and electrometer tube, with 
provision for refrigerating the cell to reduce thermionic emission. 

Chap. X] 



In both Fig. 10 and Fig. 11 is shown a convenient switching 
arrangement which provides the possibility of using any one 
of three grid resistors, or of floating the grid for rate-of-drift 
measurements. This design requires no other support for 
the sensitive grid lead than that furnished by the tube itself. 
Should other support be necessary, only the best insulating 
materials should be used. Amber or fused quartz is recom- 
mended. All insulating surfaces, including the photoelectric 
cell and the exterior of the amplifying tube, must be kept 
free from grease or dirt and should be handled as little as 
possible. They may be cleaned by swabbing with cotton 
moistened with 95 per cent ethyl alcohol. 

All parts of the circuit should be enclosed in metallic 
shielding. The various control resistances, the galvanom- 
eter shunt, and necessary meters may be mounted in a 
metal or metal-lined box such as that shown in Fig. 12. 

Ayrton shunt 


aluminum and 
over front 
not shown} 


cables to 
photocell ^, metal- \| 
shielded batter 
boxes and galvanometer "~~~ 

Fig. 12. Control box for electrometer tube circuit. 

current switch \ coarse «jioV 

fine adjustment 


Placing the battery in a shielded box is also advisable. 
Temperature insulation on this box to cut down drifts will 
be helpful. The shielded multiconductor cables available 
from radio supply houses will be suitable for connections 


between these various units, or the necessary wires may be 
pulled through a flexible metal hose. 27 

Whenever possible, all connections should be soldered, 
using only rosin flux. A flux made by dissolving rosin in 
alcohol will be found convenient; it may be applied with a 
brush. It is usually not necessary to solder the wires to the 
pins of the tube base, but a socket which grips the pins 
tightly must be used. The soldering of connections to the 
battery terminals is particularly important. 

The batteries should be kept in first-class condition; the 
tops should be kept clean and dry. They should be charged 
at a moderate rate. It will be found advantageous to oper- 
ate them only on the middle portion of the discharge curve. 

The resistances, rheostats, and potentiometers should be 
of good quality. Wire-wound resistors may be obtained 
from any radio supply house. General Radio Type 371 
potentiometers have been found satisfactory. Ohmite 
model H and model J units are also to be recommended. 
The inexpensive small wire-wound potentiometers of the 
type used in radios are not above suspicion, but Clarostat 
controls have been found entirely satisfactory by some 
workers. The high resistances to be used in the control-grid 
circuit are supplied commercially with values up to 10 12 
ohms. 28 These resistors have proved so satisfactory that it 
is not worth while for the worker to attempt to make his own. 

In certain applications, in which portability is required or 
vibration makes it difficult to use a sensitive galvanometer, 
an additional stage of amplification may be necessary. 
DuBridge 29 used a 112A triode for the second stage and 
obtained an over-all current amplification of 4 X 10 7 . The 
indicator was a microammeter. The design of the second 
stage is not critical, and almost any tube with a high mutual 
conductance is satisfactory. However, for most laboratory 

27 Such hose is obtainable from the American Metal Hose Company, Water- 
bury, Connecticut. 

28 These resistors may be purchased from the S. S. White Dental Manu- 
facturing Company, Industrial Division, 10 East 40th Street, New York City. 

™ DuBridge, L. A., Phys. Rev., 37, 392 (1931). 

Chap. X] 



applications a single electrometer tube and galvanometer is 
preferable. A galvanometer with a sensitivity of 5 X 10~ 10 
ampere/mm at 1 m is simple and foolproof, and amply- 
sensitive to show the inherent unavoidable fluctuations of 
the amplifier. 

Other low-grid-current tubes. Several workers have 
pointed out that many commercial radio tubes, when oper- 
ated at low potentials, have greatly reduced grid currents, and 

Fig. 13. Plate- and grid-current characteristics of RCA 38. 

frequently by selection a tube may be found which has a grid 
current as low as 10 -12 ampere. After a careful investiga- 
tion Johnson found 30 that the RCA 38 pentode had good 
characteristics in this respect. Dunning 31 has recommended 
the Western Electric 259-B screen grid tube. MacDonald 32 
has found the RCA 22 very satisfactory at low voltages and 
has given complete data on its characteristics. 

30 Johnson, E. A. and A. G., Phys. Rev., 50, 170 (1936). 

Johnson, E. A., and Neitzert, C, Rev. Sci. Instruments, 5, 196 (1934). 

31 Dunning, J. R., Rev. Sci. Instruments, 5, 387 (1934). 
82 MacDonald, P. A., Physics, 7, 265 (1936). 


The grid current of a tube is often the most important 
characteristic in choosing a tube to amplify small currents or 
small voltage. The following procedure is convenient in 
obtaining data for plotting the grid-current curve : The plate 
current is plotted as a function of the grid voltage in the 
usual way, first with no resistance in the grid circuit and 
then with a high resistance in series. The horizontal differ- 
ence between the two curves represents the potential drop 
in the resistance due to the grid current, and by dividing this 
voltage difference by the value of the resistor, the grid 
current in amperes is found. A shielded box is necessary. 

Characteristic curves for the RCA 38 pentode from John- 
son's data are shown in Fig. 13. This probably represents 
the results for a selected tube. From the slope of the lower 
curve it is found that the grid resistance at — 2 volts is 
10 12 ohms. At equilibrium potential this is reduced to 
5 X 10 10 ohms. 

Gabus and Pool 33 found that the HCA 954 acorn pentode 
can be operated in such a way as to have a very low grid 
current and correspondingly high input resistance. The 
No. 3 grid, normally used as the suppressor, is used as the 
control grid. Owing to the peculiar construction of the 
tube, this grid is very well insulated. The No. 1 grid, 
normally the control grid, is connected to the cathode. 
The No. 2 grid, normally the screen grid, is operated at 
a positive potential and acts as in the electrometer tube 
to protect the control grid from positive ions emitted by 
the cathode. 

A particular 954 tube tested under conditions similar to 
those recommended by Gabus and Pool had the following 
characteristics: Heater voltage, 4 volts; grid No. 1 connected 
to cathode; grid No. 2 at + 13.5 volts; grid No. 3 at — 4 
volts; plate at + 6 volts; plate current, 60 microamperes; 
mutual conductance, 100 microamperes/volt; plate resist- 
ance, 35,000 ohms; amplification factor, 3.5; grid current, 
4 X 10~ 13 ampere. 

83 Gabus, G. H., and Pool, M. L., Rev. Sci. Instruments, 8, 196 (1937). 

Chap. X] 



For many applications in measuring currents down to 10~~ 13 
ampere, any of the above-mentioned tubes will be found 
quite satisfactory and much less expensive than electrometer 

High-gain direct-current amplifiers. Direct-coupled 
multistage d.c. amplifiers similar to the familiar a.c. ampli- 
fiers have been used very little because of the necessity of a 
separate power supply for each stage and because of difficul- 
ties with cumulative drift and instability. 

Horton 34 has described an ingenious circuit in which 
extremely high gain is realized by utilizing one high-mu 





Fig. 14. Horton's high-gain d.c. amplifier. 

pentode as the load resistance for another. In this way high 
amplification can be obtained without the very high plate 
supply voltage that would be necessary if a pure ohmic 
resistance were used for the load. The circuit is shown in 
Fig. 14. 

All plate potentials are supplied by a 180-volt battery, and 
the insulation between the heater and cathode of these tubes 
is sufficiently good that a common heater battery may be 
used. The over-all mutual conductance of this three-tube 
amplifier is about 4.5 mhos. If the amplifier is to work into 
a high-impedance load, such as a cathode-ray oscillograph, 
the type 89 output tube may be omitted and the load cgp- 

34 Horton, J. W., Frank. Inst., J., 216, 749 (1933). 


nected directly between the cathode of the second pentode 
and the 90-volt tap of the plate battery. In this case the 
voltage amplification is about 2500. Because of the capacity 
shunt through the heaters, the circuit has a poor response for 
high audio frequencies. In a similar circuit described by 
Schmitt, 35 in which separate heater batteries are used, this 
limitation is largely removed. Circuits of this type have 
found special application in the measurement of physio- 
logical potentials in which the resistance in the input circuit 
is low. Since the 77 tube (and similar high-gain pentodes) 
have a rather low grid-to-filament resistance, it is advisable 
to add a preliminary stage when measuring photoelectric 
currents. This would consist of an electrometer tube or 
other low-grid-current tube, operated from separate bat- 
teries. With such a combination a stable current amplifica- 
tion of 10 10 is possible, and the indicating instrument may 
be a milliammeter. 

Alternating-current amplifiers. One great advantage of 
an a.c. amplifier is that all the tubes of a multistage amplifier 
may be operated from a common power supply. The signal 
is transmitted from one stage to the next by means of a 
condenser or transformer, which, while providing insulation 
against steady potentials, offers a low impedance path for 
a.c. signals. Since only rapid variations are passed along, 
such an amplifier is insensitive to drifts in the plate current 
of any tube, and to gradual changes in the supply voltage. 
The characteristics of a.c. amplifiers have been treated so 
extensively elsewhere 36 that they will be considered only 
briefly here. The discussion will be limited to resistance- 
capacity coupling. 

35 Schmitt, O. H. A., Rev. Sci. Instruments, 4, 661 (1933). 

36 Chaffee, E. L., Theory of Thermionic Vacuum Tubes. New York: 
McGraw-Hill Book Company, 1933. 

Glasgow, R. S., Principles of Radio Engineering. New York: McGraw-Hill 
Book Company, 1936. 

Henney, K., Radio Engineering Handbook, Second Edition. New York: 
McGraw-Hill Book Company, 1935. 

The Radio Tube Manual, issued by the RCA Manufacturing Company, 
Harrison, New Jersey, contains much practical information. 

Chap. X] 



Fig. 15 shows a photoelectric cell and three-stage amplifier 
for use with modulated light intensities in the audio-fre- 
quency range. The power is supplied from the 110- volt a.c. 

n0 6C£ 

Rt = 5oo,ooo ohms 

R 2 = 1 megohm 

R 3 = 2.50.000 ohms 

R+ = 5o,000 ohms 

R 5 = 5 ooo ohms 

R 6 = ohms 

R 7 = 4 ooo ohms 

Ra = 25, ooo ohms 

R. 9 = 4oo ohms 

R 10 = 5oo,ooo ohms volume control 

R lt = 5o,ooo potentiometer 

110 v A.C. 
Ci=.o2-mf paper 4oovolt 

(mica better in l*i stacfe) 
C a = l-mf paper 2oo volt a 
C 3 = 5-mf electrolytic 25 volt 
C4 = 2-mf electrolytic 450 volt 
C 5 = 4-mf electrolytic 45o volt 
C 6 = 25mf electrolytic 25 volt 
C 7 = 8-mf electrolytic 45o volt 
Li = 12. henries, 80 m a. 

Fig. 15. Three-stage a.c. amplifier for photocurrents from a modulated or 
interrupted light source. 

line, rectified and filtered in the usual manner to provide the 
necessary d.c. voltages. 

In order for the coupling condensers d to perform their 
function of transmitting the voltage variation of the plate 
of one tube to the grid of the next, their impedance must be 
small compared with the associated grid resistor Ri. The 


reactance of a condenser in ohms is given by the well-known 
formula X = 1/(2tt/C), where / is the frequency in cycles per 
second and C is the capacity in farads. A simple computa- 
tion shows that when C\ = 0.02 microfarad, the reactance 
becomes equal to 500,000 ohms, the value of the grid re- 
sistor R h at a frequency of 16 cycles. This, then, is roughly 
the low-frequency cutoff of the above amplifier. 

The high-frequency cutoff is determined by the tube 
capacities shunting the plate-load resistors and grid resistors. 
At high frequencies, the coupling condenser has negligible 
impedance, so that the coupling resistors may be considered 
to be in parallel. Considering first the photocell circuit, the 
resistors R 2 and R\ have a resistance in parallel of 333,000 
ohms. The capacity shunting this is the sum of the capacity 
of the photocell, the connections, and the dynamic input 
capacity of the first tube. The dynamic input capacity is 
given by the expression 

C = Co + C of + [1 + l iR l /{R v + Ri)ip 0Vi 

where C is the capacity of the photoelectric cell and connec- 
tions, C gf is the grid-to-cathode capacity of the tube, C gP is 
the grid-to-plate capacity, jjl is the amplification factor of the 
tube, R P is the plate resistance of the tube, and R t is the 
plate-load resistance. In a screen-grid tube, such as the 
6C6, the grid-to-plate capacity is very small, and the last 
term is therefore unimportant. The grid-to-cathode ca- 
pacity is about 6 micro-microfarads. The total capacity 
may be roughly 15 micro-microfarads, which has a reactance 
equal to 333,000 ohms at a frequency of 34,000 cycles. This 
is the high-frequency cutoff for the input circuit. If a gas- 
filled cell is used, its lower response at high frequencies must 
of course be taken into account. 

The situation in the coupling between two stages follows 
the same general reasoning. In place of the capacity of the 
photocell there is the plate-to-cathode capacity of the tube 
delivering the signal to the coupling network. Also, the 
dynamic plate resistance of the tube is in parallel with the 


resistors of the coupling network. The high-frequency 
cutoff may be regulated by attaching a shunting condenser 
to either the plate-load resistor or the grid resistor. This 
has been analyzed by Johnson, 37 who showed that by proper 
choice of coupling and shunting condensers, an amplifier 
may be fairly sharply tuned to a single frequency. 

The necessary negative voltage for the grid of each tube is 
obtained by a self-bias resistor in the cathode lead. If the 
a.c. variations in plate current are allowed to pass through 
this resistor, the grid bias will be altered in such a direction 
as to produce a serious degenerative effect, and the ampli- 
fication is much reduced. Consequently the bias resistor is 
shunted by a by-pass condenser, which must provide a low- 
impedance path for the a.c. component of the plate current. 
To keep the impedance less than that of the bias resistor at 
the lowest frequencies amplified, a rather large capacitance 
is sometimes required, but inexpensive low-voltage electro- 
lytic condensers are available for the purpose. 

Some care is necessary to prevent regenerative coupling 
of the first and last stages through the common power 
supply, the result of which is to produce a type of oscillation 
known as motorboating. The remedy is a decoupling filter 
of the resistance-capacity type in voltage lead to the photo- 
cell, and to the screen grid and plate of each tube except the 
power tubes. The principle to be followed is that the con- 
denser in each filter unit should offer an impedance to the 
signal about 10 times less than its associated resistor at the 
lowest frequency passed by the amplifier. This is equivalent 
to the requirement that the time constant RC of each filter 
unit be 10 times longer than the time constant of the grid 
coupling condenser and resistor. For the amplifier shown in 
Fig. 15 the time constant of the coupling circuits is 0.01 
second throughout. The time constant of the plate and 
screen-grid decoupling filters of the 6C6 tube (E 6 (7 4 ) is 
0.2 second, thus more than meeting the requirement. It is 
worth while to have a good margin of safety in the filters for 

37 Johnson, E. A., Physics, 7, 130 (1936). 


the photocell and first tube, which are, of course, most 
sensitive to feedbacks. All such filters also act to eliminate 
any ripple in the rectified d.c. voltage from the power pack. 
Shielding of the photocell and first tube is necessary to 
eliminate undesirable pickup. 

The above-described amplifier is illustrative of the general 
design of a.c. amplifiers. For detecting very small fight 
intensities, additional precautions must be taken, with 
special attention given to the first stage. A grid resistor of 
high value is used and the cell connected directly to the grid 
without a coupling condenser. A tube with low grid current 
is desirable, but an electrometer tube is ruled out because of 
its very low amplification factor and its sensitivity to micro- 
phonic disturbances. The RCA 38 or the Western Electric 
259B operated at reduced voltages is probably the most 
satisfactory tube for the first stage. Careful shielding is of 
course necessary. Each tube with its coupling resistors and 
condenser should be in a separate metal compartment, with 
an adjacent enclosure for the associated bias resistor and 
decoupling filters. The tube for the output stage must be 
selected with special reference to its load. The next to the 
last tube is usually of a type intermediate between a voltage 
amplifier and a power amplifier. Johnson and Neitzert 38 
have described an amplifier for small a.c. voltages which 
used RCA 38 pentodes at reduced voltages for all stages 
except the output. Separate plate batteries are used for 
each stage, but the plate current is low enough so that small 
units can be used. They may therefore be placed in the 
same compartment with the tube. Difficulties with regen- 
erative coupling through the plate power supply are thus 
eliminated. Dunning, 39 in a paper on amplifiers for detect- 
ing single ionizing particles, gives many practical sugges- 

Fluctuation noises in vacuum-tube circuits. The back- 
ground fluctuations of the currents in vacuum tubes which 

38 Johnson, E. A., and Neitzert, C, Rev. Sci. Instruments, 5, 196 (1934). 

39 Dunning, J. R., Rev. Sci. Instruments, 5, 387 (1934). 


determine the smallest signal that may be detected have 
long gone by the name of "noise," a term which is convenient 
even when there is no conversion of the currents into sound. 
When extraneous disturbances due to vibration, poor shield- 
ing, poor insulation, poor connections, and the like are 
eliminated by suitable precautions, there remain three in- 
herent sources of noise, originating in the first tube and its 
input circuit: 

1. Thermal noise in the grid resistor. 

2. Shot effect of currents in the grid circuit. 

3. Tube noise. 

It is important to the laboratory worker to be able to 
calculate the expected noise voltage in a particular circuit, 
or he may erroneously attribute the noise to an extraneous 
disturbance and waste much time in attempting to improve 
something that cannot be improved, except possibly by 
redesigning the circuit. Pearson 40 has published a discus- 
sion of the subject and has given data on low-noise tubes 
made by the Western Electric Company. 

The mean square voltage appearing across the grid resist- 
ance R as a result of thermal agitation of charge within it is 
given by 

where k is Boltzmann's constant, T is the absolute tempera- 
ture, / 2 and /i are the upper and lower limits of the band of 
frequencies passed by the amplifier, and C is the dynamic 
input capacity of the first tube. 

In case the second term in the denominator of the above 
integrand is small compared with unity for all frequencies 
between f 2 and f h as is required in order to avoid frequency 

40 Pearson, G. L., Physics, 5, 233 (1934). 
See also Johnson, E. A. and A. G., Phys. Rev., 50, 171 (1936). 
For data on RCA 38 see Johnson, E. A., and Neitzert, C., Rev. Set. Instru- 
ments, 5, 196 (1934). 


distortion, the expression for thermal noise becomes, for a 
temperature of 300°K., 41 

E} = 1.64 Xl0- 20 £(/2-/i). 

As an example, we may compute that the thermal-noise 
voltage across a resistance of 1 megohm connected to an 
amplifier which has a band width of 10,000 cycles (that is, 
the audio-frequency range) would be 13 microvolts. 

The mean square voltage due to shot fluctuations of the 
grid current is 

E - !1 di 

where e is the electronic charge and I g is the sum of the 
absolute values of the positive and negative components 
of the grid current. 

The same equation applies for the shot effect of the photo- 
current from vacuum cells. In gas-filled cells each photo- 
electron releases an average charge ix g e, where p ff is the 
gaseous amplification factor. For rough computation fx g e 
may be used as the unit of charge in the formula, but Kings- 
bury 42 found that actually the noise is somewhat greater 
than that computed on this assumption. 

Tube noise cannot be less than the thermal noise in the 
internal plate resistance. A variety of factors serve to make 
it several times this theoretical minimum. The usual way 
of rating tubes is to give the resistance which, if placed in the 
grid circuit, would give the same noise voltage in the output 
as that produced by the tube itself. Low-noise tubes have 
an equivalent resistance of from 4000 ohms to 40,000 ohms. 
Tube noise is unimportant in amplifiers designed for ampli- 
fication of photocurrents, since the resistance in the grid 
circuit is practically always much higher than these values. 

41 Boltzmann's constant k = 1.37 X 10~ 16 erg/degree. At 300 °K., or room 
temperature, kT = 0.41 X 10 -13 erg = 0.41 X 10 -20 joule. The latter unit is, 
of course, the proper one to use in connection with coulombs, volts, amperes, 
and farads. The electronic charge e == 1.6 X 10~ 19 coulomb. 

42 Kingsbury, B. A., Phys. Rev., 38, 1458 (1931). 


When measuring very small currents with an electrometer 
tube, the grid resistor is made very large, or the floating-grid 
method is used. Under these conditions, where the circuit 
itself, rather than the galvanometer, determines the speed of 
response, the equations for thermal noise and shot noise take 
on a simple form. As shown by Hafstad, 43 the following 
relations then hold : 

&T — -"77 » 



E, 2 = 


It is to be noted that the thermal noise is independent of 
resistance and is therefore a very general limitation on all 
electrometers. If C = 10 -11 farad, a practical minimum for 
the grid circuit, including photocell and connections, the 
thermal noise for 300°K. may be calculated to be 20 micro- 
volts. If R = 2 X 10 11 ohms and I g = 10~ 15 ampere, the 
computed shot noise is 1.3 microvolt, negligible in compari- 
son with thermal noise. The resultant uncertainty in a 
current measurement is 10 -16 ampere. 

However, if a resistance of 5 X 10 10 ohms or less is used 
with a 6-second galvanometer, the galvanometer limits the 
speed of response, an effect equivalent to increasing the 
capacity. The observed noise voltage is then reduced to 
about 8 to 10 microvolts. A large capacity in the grid 
circuit would also reduce the noise voltage, but the deflection 
time would be increased by just enough to make the precision 
remain the same. 

When the tube is operated with the grid at the equilibrium 
potential, R is the input resistance of the tube itself, about 
10 14 ohms, and I g has two components, each of 10 -15 ampere. 
The computed shot noise is then 40 microvolts. The thermal 
noise is of course unchanged from the previous case. With 
a total noise voltage of V40 2 + 20 2 = 45 microvolts, the 
uncertainty in any measurement of charge then amounts to 

43 Hafstad, L. R., Phys. Rev., U> 201 (1933). 


4 X 10 -16 coulomb, or, for a collection time of 1 minute, an 
uncertainty in the current of 7 X 10~ 18 ampere. In this 
case it is advantageous to have a low capacity. These are 
theoretical limits, but with reasonable care they may be 
realized in practice. 

Applications of Photoelectric Cells 

General remarks on photoelectric photometry. The most 
important laboratory use of the photoelectric cell is in 
photometry. Used properly, it is capable of giving results 
of high precision. It should be remembered, however, that 
very few cells have a constant sensitivity. Both the abso- 
lute sensitivity and the color response may change with 
time. For this reason, in careful measurements the photo- 
electric cell should be called upon only to make comparisons 
between a standard and an unknown source. Some cells are 
subject to fatigue when exposed to a bright light. If the 
linear relation between current and light intensity is to be 
depended on over a range of more than two- or threefold, it 
must be tested. The inverse-square law offers a convenient 

If the light is to be projected on the sensitive surface by an 
optical system, the illuminated area should not be too small. 
An out-of -focus image about 1 cm in diameter is satisfactory. 
The light from different successive sources should cover the 
same area as nearly as possible. This reduces errors from 
local variations in sensitivity on the surface. 

There are three general methods in photoelectric photom- 

1. Substitution method. The cell is exposed alternately 
to the standard source and to the unknown, and the relative 
deflections noted. This is the simplest and most direct 
method and is capable of giving excellent results. Of course 
it depends on the linearity of the cell and the current- 
measuring instrument. 

2. Balanced-cell method. Two cells are connected in 
opposition, and the intensity of either the standard or the 


unknown is reduced by suitable means until there is a bal- 
ance. Although the method can be made very sensitive to 
unbalance, there is a fundamental objection to depending on 
the constancy of the two cells over any extended period. 

3. Flicker method. The two sources shine alternately in 
rapid succession on the cell. The intensity of the brighter is 
cut down by a suitable intensity reducer until the flicker is 
zero. The amplifier may be of the a.c. type. A 6E5 tuning 
indicator ("magic eye") makes a very satisfactory detector 
of the minimum. 44 This is an excellent method, because the 
linearity of the cell and amplifier is not an issue. 

When used in heterochromatic photometry, the photo- 
electric cell may give a judgment of relative brightness 
entirely different from that given by the eye, owing to the 
difference in spectral response. A cell with a spectral- 
response curve like that of the eye must be used if the visual 
standards of brightness are to be carried over without 
modification. The Westinghouse Photox cell meets this 
requirement very closely without the addition of any filters. 
The Weston Electrical Instrument Company supplies a suit- 
able filter for use with the Photronic cell when a response 
curve like that of the eye is required. 

For certain kinds of colorimetric measurement, the relative 
intensity through filters of various colors can give much 
useful information. Examples are the whiteness of paper 45 
and the color temperature of a lamp 46 or a star. 47 However, 
such results are so dependent on the particular cell and 
particular filters used that they can be relied on only if 
continually checked against some kind of standard. The 

44 Garman, R. L., Rev. Sci. Instruments, 8, 327 (1937). 
Waller, L. C., RCA Review, l,V.l (1937). 

45 Davis, M. N., Paper Trade Journal, July 4, 1935, page 36. The commercial 
instrument, known as a reflectance meter, is available from the General Elec- 
tric Company. 

46 Campbell, N. R., and Ritchie, Dorothy, Photoelectric Cells, page 214. 
New York: Isaac Pitman and Sons, 1934. 

47 Stebbins, J., and Whitford, A. E., Astrophys. J., 84, 253 (1936); Mi 
Wilson Contr. No. 547. 


infrared transparency of practically all glass and gelatin 
filters mentioned on page 402 must be borne in mind. 

Spectrophotometry. The first requisite in photoelectric 
spectrophotometry is a good monochromator. A double 
monochromator is to be preferred because of the greatly 
reduced stray light of other wave lengths than the one being 
used. 48 

The principal spectrophotometric measurements made in 
the laboratory involve spectral-radiation curves of luminous 
sources, spectral-transmission curves, and spectral-reflectance 
curves. For spectral-radiation curves, the thermocouple is 
preferable to the photoelectric cell because it measures 
energy directly. If, however, the intensity is too low for the 
thermocouple, a photoelectric cell may be used, provided its 
spectral-sensitivity curve is accurately known, so that 
corrections to get the energy may be applied for each wave 
length. The spectral-sensitivity curve for the cell can be 
determined by comparison with a thermocouple, using a 
bright light as the source for the monochromator. This is 
the general method for obtaining the curves shown in Fig. 1. 

Spectral-transmission curves of filters, for example, may 
be very easily determined with a monochromator and a 
photoelectric cell. The intensity with the filter in and out 
of the beam gives the percentage transmission for each wave 
length. If the cell cannot be placed close enough to the slit 
to receive all of the exit beam, an additional lens is necessary. 
A minor adaptation enables absorption cells to be introduced 
into the beam for investigating the spectral-absorption 
curves of solutions. 49 Measurements of specular reflecting 

48 Preston, J. S., Journ. Sci. Instruments, 13, 368 (1936), has described a 
simple method of eliminating errors due to stray light in a monochromator. 
Shutters which may occult half of the slit length are fitted to the entrance and 
exit slit. A reading is taken with the shutters set to let the light go through. 
Then one shutter is reversed, and any light coming through is stray light, for 
which appropriate corrections may be made. 

49 Hogness, T. R., Zscheile, F. P., and Sidwell, A. E., J. Phys. Chem., 41, 379 

Zscheile, F. P., Hogness, T. R., and Young, J. F., J. Phys. Chem., 38, 1 


power may be carried out in the same way, except that it 
will, of coarse, be necessary to move the photocell when 
shifting from the direct to the reflected beam. 

For determinations of diffuse reflectance it is customary to 
illuminate the object at 45° incidence and observe along a 
line normal to the surface. The reflectance is measured 
relative to that of some standard material, such as mag- 
nesium carbonate. The fluorescence of certain materials 
causes difficulty, which can be eliminated only by having the 
reflection occur before the light beam passes through the 

All of the above-mentioned methods of spectrophotometry 
involve the linear response of the photocell and the current- 
measuring instrument. In practical cases this is not likely to 
cause appreciable error, but in precision work the linearity 
must be investigated. Mention should be made of the Hardy 
automatic recording spectrophotometer, 50 which works on 
the flicker principle and is thus independent of the cell and 
amplifier characteristics and variations in light intensity. 
This instrument is adaptable to many kinds of measure- 
ments and gives very accurate results. It is manufactured 
commercially by the General Electric Company. 

Densitometers. In photographic photometry it is neces- 
sary to have some means of measuring the density of the 
photographic deposit. The photoelectric cell has found wide 
use as the light-sensitive unit of objective densitometers. 
Since photographic photometry consists of interpolation 
between standards whose intensity ratio is known, spec- 
tral sensitivity and exact linear response are not crucial 

For sensitometry of photographic materials, when an area 
several millimeters square is of uniform density, good results 
can be obtained by a very simple arrangement without any 
optical system at all. An automobile headlight is mounted 

50 Hardy, A. C, /. 0. S. A., 18, 96 (1928), 25, 305 (1935). An instrument 
working on a similar principle is described by Sharp, C. EL, and Eckweiler, 
H. J., /. 0. S. A., 23, 246 (1933). 



[Chap. X 

12 to 18 inches above a flat opaque screen with a rectangular 
aperture of the desired size cut in it. A photovoltaic cell is 
mounted directly behind the aperture and connected to a 
galvanometer. The plate is best laid on the aperture emul- 
sion side down to reduce scattering effects. Miss Mohler 
and Miss Taylor 51 have described a reflection densitometer 
almost equally simple and easy to construct. Difficulties 
with stray light, and the effect of scattering in the emulsion 
will, however, be reduced if a simple projection system is 

For many problems it is desired to know the density on a 
very small area of the plate — hence the term micro-pho- 
tometer for densitometers which are designed for such 


plate to be 


cell enclosed 
•n a light- 
proof hoodi 

condensing' lenses 
focusing an image 
of the filament ont 
•microscope objective' 



focusing an 

image of the 

slit on the 

plate to be measured 

Fig. 16. Optical system of Lange's microphotometer. 

applications. Lange 52 and Milligan 53 have described micro- 
photometers using photovoltaic cells. The optical system 
of the Lange instrument is shown in Fig. 16. The con- 
densing lens forms an image of the lamp filament in the 
principal plane of the microscope objective, which is ad- 
justed to give a sharp image of the slit in the plane of the 
emulsion. The projected image of the slit may be made as 
narrow as 0.01 mm. The instrument is capable of giving 
rather high resolution. 

51 Mohler, Nora M., and Taylor, Delia Ann, /. 0. S. A., 26, 386 (1936). 

52 Lange, B., Zeits.f. techn. Physik, 13, 600 (1932). 

53 Milligan, W. 0., Rev. Sci. Instruments, 4, 496 (1933). 



The optical design and other features of microphotometers 
have been reviewed by Harrison. 54 To eliminate errors 
caused by scattered light, it is customary to use at least two 
slits in high-resolution instruments. In order to obtain 
sufficient sensitivity with reasonable speed of response, it is 
necessary to use an emission-type cell and amplifier. For the 
amplifier an RCA 38 pentode, connected as in Fig. 5, is satis- 
factory. 55 Suitable operating conditions and circuit con- 
stants are as follows : Heater, 6 volts from storage battery. 
Control grid, — 1.5 volts; screen grid, + 6 volts; plate, 
+ 12 volts; all from small radio batteries. Plate current, 
60 microamperes. Mutual conductance, 150 microamperes/ 
volt. Grid resistor, 10 8 to 10 9 ohms. Galvanometer, 5 X 
10~ 8 ampere/mm. A cesium oxide vacuum photocell is 

Measurements of the transmission of solutions as a func- 
tion of depth and concentration can also be made advan- 
tageously by photoelectric methods. Many procedures for 
chemical analysis which formerly used visual methods are 
now carried out with increased precision by employing a 
photoelectric cell. 56 

The light absorption of a solution is a logarithmic function 
of the concentration. Over a certain range the density of a 
photographic plate is a logarithmic function of the intensity. 
It would therefore be convenient to have an amplifier with a 
logarithmic response in order to make a direct-reading 
densitometer. Hunt 57 found that a remote cutoff tube such 
as the type 78 can be made to have an accurately logarithmic 
response over a voltage range of about tenfold. By using 
three tubes in cascade, he was able to extend this range to 
over a thousandfold. This arrangement is for alternating- 

54 Harrison, G. R., J. 0. S. A. and Rev. Sci. Instruments, 19, 267 (1929); 
/. 0. S. A., 24, 59 (1934). 

55 Kron, G. E., private communication. 

66 Withrow, R. B., Shrewsbury, C. L., and Krayhill, H. L., Indust. & Engin. 
Chem. (Analytical Edition), 8, 214(1936). 

Strafford, N., Analyst, 61, 170 (1936). 

67 Hunt, F. W., Rev. Sci. Instruments, 4, 672 (1933). See also Ballantine, S., 
Electronics, 1, 472 (1931). 


current voltages only. Muller and Kinney 58 have applied" 
this principle to measuring the concentration of solutions. 
They report that the open circuit e.m.f . of a Weston Pho- 
tronic cell is also a logarithmic function of intensity. 

Amplification of small galvanometer deflections. The 
ultimate sensitivity of a moving-coil galvanometer is set by 
the Brownian movement. Because of the limitations of the 
optical system it is difficult to realize this limit in a galva- 
nometer with a reasonably short period. Moll and Burger 59 
described a thermo-relay in which a rotation of the coil of 

divided { ' "" ~** . 

photovoltaic „. to secondary galvanometer 

cell disk 

image of slot 


lens focusing* 
lens focusing ^ an image of 

a " »»r»age of the slot primary 

the light source on the galvanometer 

on the 5 galvanometer photocell ga.vanomeier 


Fig. 17. Amplification of small galvanometer deflections. 

only a few seconds of arc was amplified and read on a second 
galvanometer. More recently, amplification schemes using 
photoelectric cells 60 have come into use, with some improve- 
ment in the speed of response. 

The simplest method of amplification involves a balanced 
photovoltaic-cell arrangement. The sensitive disk of a 
Weston Photronic cell is removed from the case. This may 
be done by unscrewing the back, warming gently, if neces- 
sary, to soften the pitch seal. The conducting layer on the 

58 Milller, R. H., and Kinney, G. F., J. 0. S. A., 25, 342 (1935). 

59 MoU, W. J. H., and Burger, H. C, Phil. Mag., 50, 6211 (1925). 

60 Jones, R. V., Journ. Sci. Instruments, 11, 302 (1934). 

Moss, E. B., Journ. Sci. Instruments, 12, 141 (1935). This is a general 
review of the subject. 
See also Taylor, A. H., Rev. Sci. Instruments, 8, 124 (1937). 


top surface is divided into two parts by scratching a line 
along a diameter of the disk with a sharp instrument. Any 
loose particles must be brushed away. If this process is 
carried out properly, the two halves of the top conducting 
layer will be insulated from each other. Fig. 17 shows the 
optical arrangements. An automobile headlight bulb is 
focused on the mirror of the primary galvanometer by a 
simple lens, in the plane of which is a rectangular slot. The 
galvanometer lens forms an image of the illuminated slot on 
the divided disk of the photovoltaic cell. The two halves of 
the top conducting layer are connected to the secondary 
galvanometer, which indicates the difference in the illumi- 
nation on the two sides of the center line. A distance of 
1 or 2 feet between the photocell and the primary galva- 
nometer is sufficient. The secondary galvanometer may 
be placed wherever it is convenient. 

A current amplification of 200 is easily obtained with such 
an arrangement, which is sufficient to make the Brownian 
motion quite noticeable. Another application would be to 
couple two short-period, low-sensitivity galvanometers by 
such a device to form a high-sensitivity combination with 
rapid response. The over-all linearity is, of course, open to 
suspicion and must be tested. 

Thyratrons. The introduction of a gas into a hot-cathode 
tube greatly increases the power it can handle, owing to the 
neutralization of space charge by the positive ions formed. 
However, the grid-control characteristic is quite different 
from that of a high-vacuum tube. With the grid sufficiently 
negative, the tube is nonconducting. At a certain critical 
grid potential, whose value depends on the plate voltage, the 
discharge starts, and the tube is said to "fire." The grid 
then loses control of the plate current because of a sheath of 
positive ions around it. The discharge can be stopped only 
by removing the plate voltage. The voltage drop in the 
tube is practically independent of current and amounts to 
about 15 volts. The current must be limited by resistance 
in the external circuit to a safe value. 



[Chap. X 

The name thyratron 61 was first applied to these tubes only 
by the General Electric Company, but is coming into general 
use. Other terms sometimes used are gaseous triode and 
grid-controlled rectifier. The Westinghouse Lamp Company 
and the RCA Manufacturing Company also make tubes of 
this type. The gases used are argon and mercury vapor. 
Mercury-vapor tubes have the disadvantage, which is, how- 
ever, usually not serious, that the characteristics are a func- 
tion of the ambient temperature. Argon tubes are not 
subject to this variation, but are limited in their power- 
handling ability. The General Electric FG-81, the largest 
argon-filled tube, is rated at 0.5 ampere maximum average 
plate current and 180 volts maximum plate voltage. 

In most applications gaseous triodes are used with an 
alternating plate voltage. During the negative half of the 

cycle the grid regains control. 
The output is then a pulsat- 
ing direct current. Tubes may 
be used in pairs if full-wave 
operation is desired. 

The simplest application is 
as a relay. As compared to a 
mechanical relay, the gaseous 
triode requires much less power 
to operate, is faster and quieter, 
and has no contacts to pit, 
wear, or stick. Fig. 18 shows 
a gaseous triode used as the 
relay in a thermostat controlled by a mercury thermometer. 
A load of 2 kilowatts may be switched on and off with a 
current at the mercury contact of a few microamperes. The 
mercury is thus protected from contamination due to 

This particular application is merely illustrative. The 
control may be exercised by a photoelectric cell, and the load 

61 For many details about characteristics and uses, see Hull, A. W., Gen. El. 
Rev., 32, 213, 390 (1929); Physics, 4, 66 (1933). 

AC line 

Fig. 18. Use of thyratron as a 
relay in a mercury-controlled 

Chap. X] 



may be a light, a motor, or a magnet. If response to even 
smaller grid power is required, a four-electrode tube known 
as a shield-grid thyratron is available. 62 

Continuous variation of the average plate current between 
zero and full value may be obtained by a phase-shift circuit, 
shown in Fig. 19. The phase of the alternating grid voltage 
is varied by the relative values of C and R. If R is very 
large, the grid voltage is 180° out of phase with the plate 
voltage, and the tube is always nonconducting. If R is zero 
or very small, the grid volt- 
age is in phase with the plate 
voltage, and the tube fires at 
the beginning of each positive 
half-cycle and carries the maxi- 
mum current. At an inter- 
mediate value, the grid volt- 
age will reach the critical po- 
tential at some time during 
the positive half-cycle, and 
the tube conducts for the re- 
mainder of the half -cycle. A 
smooth and fairly linear variation of the average current 
is thus possible. The resistance R may be a vacuum tube 
or a photoelectric cell. 

In Fig. 20 is shown a phase-shift circuit for maintaining a 
constant temperature in a furnace. 63 In a test, the tempera- 
ture of a furnace at approximately 880 °C. was held constant 
within 0.06°C. The temperature is measured by a thermo- 
couple. (A resistance thermometer would do as well.) The 
potentiometer is set to balance at the desired constant tem- 
perature. An automobile headlight bulb is imaged on the 
galvanometer mirror by the lens L\. The galvanometer 
lens L 2 forms an image of Li on a V-shaped slot in front of 
the photocell. The amount of light reaching the cell (which 
should be of the gas-filled type) determines the current 

62 Livingston, O. W., and Maser, H. T., Electronics, April, 1934. 

63 Zabel, R. M., and Hancox, R. R., Rev. Sci. Instruments, 5, 28 (1934). 

Fig. 19. 

A.C. line 

Phase-shift circuit for con- 
trol of thyratron. 



[Chap. X 

through the thyratron and thus regulates the furnace tem- 
perature. Ri is in parallel with the regulator circuit and 
carries most of the heater current, thus making it possible to 
use a smaller tube. R 2 limits the current through the tube to 
its rated maximum value. C may be a 200 micro-microfarad 
variable condenser. 

This arrangement is an example of a type of control mecha- 
nism that will be found generally useful in the laboratory. 

Fig. 20. Constant temperature regulator using a photocell in a phase-shift 


It differs from the simple on-and-off control provided by 
relays in that the correcting influence approaches zero gradu- 
ally as the error diminishes, and " hunting" is thereby 
eliminated. The same principle is applicable to devices for 
maintaining constant speed, constant current, or constant 
voltage. With two lights and two photocells, it may be used 
to control a motor which will balance a bridge or a scale, or 
perform any other " centering" operation. 64 

64 For further applications see Henney, K., Electron Tubes in Industry, 
Second Edition. New York: McGraw-Hill Book Company, 1937. 


Photography in the Laboratory 1 

TN this chapter we will treat of photography and the 
-I photographic procedures used in making and recording 
observations in experimental science and especially in record- 
ing spectra. 

Comparison of the sensitivities of the eye with the photo- 
graphic emulsion. The relative visibilities of the various 


400 450 500 . 550 600 650 700 750 

wave length in xy\u 

Fig. 1. 

colors of the spectrum are shown in Fig. 1. To illustrate the 
differences of response to various wave lengths between the 
eye and the photographic plate, these visibility curves are 
to be compared with the sensitivity curves for ordinary, 
orthochromatic, and panchromatic emulsions that are shown 
in the top section of Fig. 2. 

1 1 wish to acknowledge the use of material from the following sources for 
the preparation of this chapter: 

Elementary Photographic Chemistry. Rochester: Eastman Kodak Com- 
pany, 1931. 

Fowle, F. E., Smithsonian Physical Tables. Washington: The Smithsonian 
Institution, 1934. 

Hardy, A. C, and Perrin, F. H., The Principles of Optics (chapter on photog- 
raphy). New York: McGraw-Hill Book Company, 1932. 

Neblette, C. B., Photography. New York: D. Van Nostrand Company, 


ordinary \ 

panchromatic "^ * 

+ 5- 

300 400 500 . . 600 700 
waive length in m a 

exposed to sunli g ht 


Eastman infYared sensitive plate 
type 1HJ 

Kodak infrared film 

300 -too 

500 600 



Fig. 2. 

Chap. XI] 



Wedge spectrograms are illustrated in the lower sections 
of Fig. 2. These spectrograms were taken using sunlight 
and the light of a tungsten lamp as the light source. The 
height of the shaded areas indicates the sensitivity of the 

Owing to the emission characteristics of these light sources 
and the opacity of the lenses used for violet and ultraviolet 
light, the wedge spectrograms exhibit maxima at about 
4700 A, whereas, actually, the sensitivities of the emulsions, 
as indicated by the curves at the top of Fig. 2, show the 
maxima to be in the ultraviolet. However, the wedge 
spectrograms do give an indication of the performance of 
different emulsions in the camera. 

Table I illustrates the relative sensitivities of the eye and 
the photographic plate to a line-shaped light source on a dark 
field. This table, together with the appearance of a spec- 


Color of Linear Image of Light on 

Dark Background Just Visible at a 



Blue (4500 A) . 
Green (5200l) 
Red hydrogen . 
Extreme red . . 

Time Required to Register the Line 

Photographically on a Panchromatic 


1 minute 

5 minutes 

30 hours 

17 minutes 

1 minute 

The material for this table appeared in the Scientific American a few years 

trum line, may be used to determine the approximate ex- 
posure time for a spectrum plate. 

Hurter and Driffield curves. The characteristics of 
photographic films and plates are simply represented by 
Hurter and Driffield curves 2 (designated hereafter as H and 

2 Hurter, F., and Driffield, V. C, Jour. Soc. Chem. Ind., 9, 455 (1890). 
Ferguson, W. B., The Photographic Researches of Ferdinand Hurter and Vero 
C. Driffield. Royal Photographic Society, 1920. 



[Chap. XI 

D curves). An H and D curve is shown in Fig. 3. This 
curve represents the relationship between photographic 
response and the exposure to white light which is required to 
produce this response. The response, measured by the 

region of 

10 10" to 3 io 4 io 5 

exposure (lumens/meter 2 -seconcts) 

Fig. 3. 

photographic density, A, is plotted as ordinate against the 
logarithm of the exposure, 2, as abscissa. A is defined by the 



A = log; 



in which T is the transmission of the film or plate in question. 

The shape of the H and D curve depends upon the char- 
acter of the emulsion, and also, if colored light is used, upon 
the wave length or color characteristics of the light used for 
exposure. Actual photographic materials may exhibit a 
curve differing considerably from the one shown in Fig. 3. 
For example, the curve does not always exhibit a definite 
straight segment in which A is proportional to the logarithm 
of exposure. The curve of Fig. 3 is somewhat idealized, but 
it represents the general character of the relationship be- 
tween A and log 2. 

The exposure range represented by the straight segment is, 
by definition, the latitude of the emulsion. Table II gives 
the latitude of typical emulsions, from which we see that fast 

Chap. XI] 



emulsions have the greatest latitude and lantern slides and 
process emulsions the least, the variation of latitude in these 
emulsions being eightfold. In Table II the latitude is 
expressed as the ratio of the exposure at the upper end of the 
straight segment to the exposure at the lower end. 

Latitude of Photographic Emulsions 

Photographic Material 

Latitude (ratio of exposure at limits 
of the straight segment of the A-log 2 
curve for development to Too) 

Motion-picture film: 

Extra fast and normal 












Process plates 


Lantern-slide plates 


Smithsonian Tables, page 342 (1936). 

The contrast of an emulsion, y, is, by definition, the slope 
of the straight segment of the H and D curve. If this seg- 
ment makes an angle a with the log 2 axis, 

7 = tan a. (2) 

The contrast varies with development time but tends to 
approach a limit as the development time is increased. This 
limiting value, y m , allows comparisons to be made between 
the contrast characteristics of various types of emulsions. 
Values of 7 oo for different emulsions are given in Table III. 
It will be noted that positive films, lantern slides, and process 
emulsions exhibit the most contrast, while fast emulsions 
exhibit less contrast. 

The curve shown in Fig. 4, the so-called Weber-Feckner 
curve, illustrates the variation of the subjective response of 



[Chap. XI 

the eye to field brightness. In several respects it is like an 

H and D curve. 

TABLE ill 

Relative Contrast of Photographic Emulsions 


Super-speed motion-picture film 
Par-speed motion-picture film . . 
Motion-picture positive film . . . 
Commercial orthochromatic 

Commercial panchromatic 

Ordinary commercial 

Process plates 

Lantern-slide plates 










Smithsonian Tables, page 342 (1934). 

The Weber-Feckner curve employs the logarithm of 
field brightness as abscissa, and this curve exhibits an inflec- 
tion point near which it is closely approximated by a straight 
line. The proportionality of A for an H and D curve, as 
well as the subjective response of the eye for a Weber- 
Feckner curve, to the logarithm of the "amount of light" 



io* io~ l io° io id* id 5 lo* io 5 

stimulus (field brightness in candles/meter 2 ) 

Fig. 4. After Hardy and Perrin. (See footnote 1.) 

indicates why photographs look natural. It also indicates 
why the ordinary fading type of exposure meter works as 
well as it does. 


The contrast sensitivity of the eye is related to the slope 
of the Weber-Feckner curve, and it is such that one can 
just distinguish a difference in brightness of about 2 per cent 
between contiguous uniformly illuminated fields. It is of 
interest to point out that the contrast of a field can be en- 
hanced by photography by about ninefold — threefold by the 
photographic process of taking the primary photograph and 
threefold by the printing process. Accordingly, it is possible 
to see detail in a photograph which is invisible to the unaided 

The H and D speed of a photographic material S is defined 
as 34 divided by the inertia expressed in lumen seconds per 
square meter. Q . 

S = ~ (3) 


The inertia, illustrated in Fig. 3, is the exposure repre- 
sented by the intersection of the extrapolated straight seg- 
ment of the H and D curve with the line A = 0. 

Reciprocity law. The photophysical and photochemical 
effects produced in a plate by the exposure 2 are measured 
by the density A. Equal effective exposures 2' produce 
equal densities under the condition of controlled develop- 
ment. To a degree of approximation sufficient for most 
applications (precise photometric photography excepted) 
the effective exposure is equal to the product of the illumi- 
nation on the plate, /, multiplied by the time of exposure t. 

ThuS > S' = S = J.*. (4) 

This is the so-called Bunsen and Roscoe reciprocity law. 

For greater accuracy it is necessary to replace Eq. 4 by 
more complicated expressions. One of them, Schwarz- 
schild's, 3 takes into account the difference in effectiveness 
when the emulsion is exposed a short time to a bright light 
and when it is exposed a long time to a dim light. Schwarz- 
schild's relation is given below: 

2' = I-t*. (5) 

3 Schwarzschild, K, Astrovhys. /., 11, 89 (1900). 


We may take the behavior of motion-picture positive film as 
an example to show to what extent the Bunsen and Roscoe 
law fails. For the range of illumination intensities from 1 to 
33,000 it is found that p varies from 0.68 to 1.00, the maxi- 
mum intensity being 131 lumens/m 2 and the exposure time 
varying between 18.2 hours and 2.5 X 10~ 4 second. 

Another factor to be considered in predicting the photo- 
graphic response to a given exposure is whether the illumi- 
nation is intermittent or continuous. The photographic 
emulsion is incapable of responding as completely to an 
exposure impressed as short flashes of light as to an equal 
uniform exposure. Also, everything else being equal, the 
photographic response is diminished if the time interval 
between flashes increases. 4 

The resolving power. The resolving power of a photo- 
graphic plate may be measured by the number of lines per 
millimeter which can be distinctly photographed. The 

Resolving Power of Photographic Emulsions 


Optimal Resolving Power 

Motion-picture film: 

Extra fast , 





Ordinary commercial 

Commercial orthochromatic 

Commercial panchromatic 

Process ordinary 


Process panchromatic 

Lantern slide 


Smithsonian Tables, page 343 (1934). 

4 For a more complete treatment of the failure of the reciprocity law, see 
Jones, L. A., "Measurements of Radiant Energy with Photographic Ma- 
terials," Measurement of Radiant Energy. New York: McGraw-Hill Book 
Company, 1937. 

Chap. XI] 



resolving power varies by a factor of about 2 for the ordinary- 
photographic materials. For lantern-slide plates the resolv- 
ing power is about 100/mm, but for fast motion-picture film 
it is only half as great, or about 50/mm. (See Table IV.) 

Light sources. Photographs are frequently taken with 
achromatic lenses. The flint components of these lenses will 
not transmit light at wave lengths shorter than about 3300 A. 
The transmission curve for a moving-picture projection lens 
containing a flint component is shown in Fig. 5. In practice a 






"*" ^**^*~ 




+ S 








• +} 













* J 

/ 4- 


















■*■ ^ 

s^ + 








■ c 



4 / 

















- O 








- t 









400 50O 

waive length in rrtyU 



Fig. 5. Transmission of a motion-picture projection lens. 

wave-length limit of 3800 A is often low enough to define the 
behavior of a photographic material when account is taken of 
this opacity of the lenses used and also when the ultraviolet 
emission of ordinary light sources is considered. 

The spectral energy distributions in the light of various 
tungsten lamps used for photography are given in Fig. 6. 
It will be noted that the emission is weak at short wave 
lengths. The spectral distribution of sunlight is also given 
there. As a matter of convenience for comparing the curves, 
the intensities are all set equal to 1 at 5600 A. 

The color temperatures of various sources are given in 
Table V. Of these, the photoflash lamp is of special interest. 



[Chap. XI 

It is of value for inside photography, since it eliminates the 
danger and smoke of flashlight powder and the expense of 
other suitable sources such as the electric arc. This lamp is 
an ordinary pear-shaped bulb filled with aluminum foil and 
oxygen. The foil is ignited by a "fuse," or small chemical 




500 600 

wave length in xrxju 

Fig. 6. Data for tungsten lamps supplied by the General Electric Com- 
pany, Nela Park, Cleveland, Ohio. 

flash, set off by applying 3 volts or more to the screw socket 
of the base. When two or more of the lamps are close to- 
gether, it is necessary to apply voltage to the base of only 
one of them and the others will go off "sympathetically." 
The light generated is 22 to 180 thousand lumen seconds, 
depending upon the size of the lamp. This light is emitted 
in an interval of time varying from ^V to -jt second. Half 


the total radiation is emitted in an interval of xJo- to -^io 
second. When the voltage is applied to the base, the lag 
between the time when the fuse is operated and the instant 
of maximum illumination is about •£% second, or, if the lamps 
are "sympathetically" flashed, the lag between the first and 
second flash is about 7V second. The maximum light in- 
tensity from a bulb of the size of an ordinary 75-watt tung- 
sten lamp is 4 to 5 X 10 6 lumens. 


Color Temperatures of Various Sources (Temperature 
of a Black Body Giving the Same Color of Light) 




Nernst filament 

Ordinary tungsten lamp . . . 

Photoflood lamp 

Flash powder 

Ordinary cored carbon arc. 
Photoflash lamp 

Color Temperature 



2780 to 3000 




Filters. For photographing with the microscope, in order 
to prevent excessive heating, it is necessary to use a water 
filter to absorb out heat radiation from the light emitted 
from the carbon arc. Frequently, copper chloride or sul- 
phate is added to the water in the cell to increase its infrared 
absorption. These additions do not materially attenuate 
the transmission of the cell for yellow, green, and blue light. 

On the other hand, when it is required to transmit the 
infrared and absorb the visible rays, one can use a cell con- 
taining a solution of iodine in carbon disulphide. 

Color filters for use in photography can be made up from 
solutions, or they may be purchased in the form of colored 
gelatin films, either unmounted or mounted between glass 
plates. The unmounted films are the cheapest, and they 



[Chap. XI 

generally serve as well as those that are mounted. Trans- 
missions for some of the Wratten gelatin filters 5 used in 
photography are given in Fig. 7. 


<t? B9 " • * \. % :'^B * 


ZK No.8 




G No.15 

Has '."v"''"^^^^ ■*■ 



A No. 25 






,j,- ■ i ^ '''"'"!■ -•'" ■- hhBk ^-^SmW * 




+ - 


No. 36 


No. 17 



2 B ^J ^iHh^V 

■ 1 



♦ r 


No. 83-A 

inf ra- 


1 '*^7' "■" ^ 1—. ., | 




300 400 500 600 

wave lengths in m/j 

Fig. 7. Wratten filters. 

There are several preliminary procedures to be carried out 
before the photographic emulsion is ready for exposure. 
These include focusing, sensitizing, annealing, and judging 
the correct exposure time to be given. 

5 Wratten niters are handled commercially by the Eastman Kodak Com- 
pany, Rochester, New York. See the book Wratten Light Filters, published 
by that company, price 50 cents. 


Focusing. Focusing may be accomplished by taking 
several pictures at different positions of the plate holder or 
lens. Sometimes, however, this procedure is not the most 
convenient one. For example, in focusing the plate in an 
astronomical telescope, a "knife-edge" test may be used for 
determining the focus. This is accomplished by means of an 
attachment which is substituted for the plate holder and 
which is constructed so that the knife edge comes to the same 
plane as that occupied by the emulsion when the plate holder 
is in position. The camera as a whole is then adjusted until 
the cutting of the knife into the star image results in a uni- 
form decrease in the intensity of the rays that come from 
different parts of the mirror. After this, when the plate 
holder is replaced, the emulsion will be in focus. 

In a spectrograph a strip of paper across the center of the 
telescope or collimating lens and parallel to the slit facilitates 
the determination of the focus. When the photographic 
plate or viewing glass is not in exact focus, a double image of 
the spectrum lines is obtained. 

A uranium glass plate may be used for focusing ultraviolet 
light. The fluorescence of this glass is easily visible for the 
stronger ultraviolet lines in the mercury spectrum. 

Sensitizing. Ordinary photographic plates respond to 
wave lengths from 2400 A to 5500 A, with the maximum 
sensitivity at about 3600 A. Sensitization is required 
beyond these limits. Although the photographically active 
grains themselves are sensitive for all wave lengths shorter 
than those of the visible spectrum, the emulsion becomes less 
sensitive at wave lengths below 2800 A, because of opacity 
of the gelatin in the emulsion. Although absorption by the 
gelatin is weak at 2800 A, it increases at shorter wave lengths, 
especially below 2400 A, until the gelatin is completely 
opaque at 2000 A. Schumann made the first photographic 
plates which were sensitive at wave lengths below 2000 A 
by using a very thin emulsion almost free from the opaque 
gelatin. Not only are the Schumann plates useful for photo- 
graphing in the ultraviolet spectrum, but they also serve in 



[Chap. XI 

those applications in which the lateral motion in the emulsion 
is to be rigorously avoided or in which lateral scattering of 
light must be minimized. 

The sensitivity of ordinary photographic plates may be 
increased at wave lengths below 2400 A by coating them with 
a fluorescent substance such as oil. The sensitivity can be 

300 40O 


6OO 700 8OO 


1O0O 1100 




















1 * i 








30O 400 50O 600 700 SOO 900 
wave length in mju 

1000 1100 


Fig. 8. Diagrammatic representation of the sensitivity of spectrum plates 
available from the Eastman Kodak Company. 

increased as much as four hundredfold. A few drops of 
Nujol or some other oil are spread over the surface with a 
cotton pad. After exposure and before development the oil 
is washed away with acetone. Harrison has studied the 
sensitometry of oiled plates, and he states that they may be 
used for photographic photometry. 6 

6 Harrison, G. R., J. 0. S. A., 11, 113 (1925). 


Plates are sensitized for the red and infrared with dyes. 
The diagram shown in Fig. 8 represents the characteristics of 
spectrum plates which are available commercially. 7 

Plates may be hypersensitized by bathing them in either 
an ammonia or a borax bath. The formulas for these baths 

are given below. 

Ammonia Fokmula 

Bath Temperature: 10° to 12 °C. 

Ammonia (0.91 sp. gr.) 2 cc 

Alcohol 275 cc 

Distilled water 725 cc 

Immerse 2 minutes. Do not rinse. Dry as quickly as possible 
after removing surface liquid. 

Borax Formula 

Bath Temperature: 12°C. 

Sodium chloride 0.5 g 

Borax 2 to 3 g 

Distilled water 1 liter 

Immerse 2 to 6 minutes. Do not rinse. Soak in methyl alcohol 
1 minute. Dry as quickly as possible after removing surface fluid. 

The hypersensitizing action of the baths is relatively 
greater for the slower emulsions. Treatment with the 
ammonia bath can be expected to produce 100 to 400 per 
cent increase of speed for the visible spectrum, while infrared 
plates exhibit an even greater increase in speed of 500 to 
2500 per cent. Hypersensitized plates fog rapidly at room 
temperature, so that they should be used as soon as possible 
after they are dry. However, they can be kept for a few 
days in an icebox. After they are removed from the icebox, 
they should be warmed to room temperature to avoid the 
condensation of moisture when they are loaded in the 
camera. The borax treatment is said to be best for Agfa 
plates, and the ammonia bath for Eastman plates. 

Photographic plates may also be hypersensitized by a 36- 
hour exposure to mercury vapor at ordinary temperatures. 8 

7 Mees, C. E. K, /. O. S. A., 25, 80 (1935). 

8 See Dersch, F., and Duerr, H., J. Soc. of Motion Picture Engineers, 28, 
178 (1937). The effect of exposure to the mercury vapor is said to be more 
marked after exposure than before. 


We see from the H and D curve given in Fig. 3 that a 
certain exposure is required before the linear part of the 
curve is reached. R. W. Wood introduced the procedure of 
pre-exposing the plate to a uniform illumination. 9 Although 
this allows the attainment of increased density with a given 
exposure, it does so with attendant loss of contrast. To 
avoid the Herschel effect, one should use, for the pre- 
exposure, a wave length that is redder than the light that is 
being photographed. Other investigators who have experi- 
mented with pre-exposure, notably Whipple in connection 
with stellar spectroscopy and Norman with photographic 
photometry, may be consulted for further details. 10 

Gelatin shifts. Annealing of the unexposed emulsion 
may be required for photographic plates such as spectrum 
plates and astrographic plates on which it is necessary to 
make measurements of the highest precision. Gelatin shifts 
may occur on account of strains, in the supporting gelatin 
layer, which are relieved by the developing process. Accord- 
ingly, the position of silver in the developed image may not 
coincide exactly with the position of the same silver in the 
latent image. Cooksey and Cooksey observed shifts of the 
latent image as great as 9fx. n These authors found, how- 
ever, that such shifts are reduced about fourfold by the 
following annealing procedure : The plate is wet in a neutral 
solution, washed, and then dried by absorbing the water 
from the gelatin with alcohol. 

More frequently gelatin shifts arise on account of improper 
drying. Large shifts occur around spots where the gelatin does 
not dry uniformly. To avoid these shifts it is recommended 
that, after fixing, the plate be dried in an alcohol bath, in 
which the removal of water from the gelatin is much more uni- 
form than it is when the drying is effected by evaporation. 
It is advisable not to use a spectrum plate nearer the edge 
than 1 cm, since the gelatin shifts are greatest near the edge. 

9 Wood, R. W., Astrophys. J., 27, 379 (1908). 

10 Norman, D., J. 0. S. A., 26, 407 (1936). 
Whipple, F. L., Lick Observat., Bull, No. 442. 

11 Cooksey, D., and Cooksey, C. D., Phys. Rev., 36, 80 (1930). 

Chap. XI] 



Because of their shrinking and expanding, photographic 
films are unsuitable for precise photography. The effect of 
changes in the film can, in a large part, be allowed for if 
coordinate lines (with a reseau) or, in the case of a spectrum, 
comparison spectral lines, are impressed on the plate at the 
time of exposure. 

Exposure. Photographic films and plates are ordinarily 
exposed to get proper contrast. As long as the range of 

range of illumination 
in object t/0 be 

H — H 

r s at i s factory 

>°g 10 2 

Fig. 9. 

illumination in the object to be photographed is small com- 
pared to the latitude of the emulsion, the exposure can vary 
between wide limits. As a result, the density may vary, but 
as long as the exposure falls within these limits, the contrast 
obtained will be the same. This is illustrated in Fig. 9. 

y>i ^^ 



i y 
i / 

/ range of 



exposure for detail 
in highlights 

exposure for detail 
in shadows 

Fig. 10. 

If the range of illumination is greater than the latitude as 
shown in Fig. 10, the exposure should be adjusted to get 
proper contrast where it is desired. For example, if the 


highlights are important, the emulsion should be under- 
exposed for the shadows, and if the shadows are important, 
it should be overexposed for the highlights. 

In photography, particularly artistic photography, one 
may wish to render detail in both highlights and shadows 
even where the range of illumination may involve a ratio as 
great as 3000 to 6000 between the maximum and minimum 
exposure. Although no film exhibits such a latitude (see 
Table II), this range of exposure can often be managed by 
overexposure and underdevelopment. That this procedure 
yields less contrast and greater latitude is indicated by 
Fig. 11, in which it will be noted that the projection of the 
straight segment of the A-log 2 curve representing 32 
minutes development embraces an exposure range only half 
as broad as that embraced by the curve for 2 minutes devel- 
opment. Furthermore, the regions of exposure beyond the 
straight segment do not deviate so much from a straight line 
for the case of 2 minutes development as for the case of 
32 minutes development. 

On the other hand, when a subject is "flat," such as a 
sand dune, the professional photographer takes his largest 
camera and gives the plate a short exposure and then a long 
development to get maximum contrast. The photographer 
uses the large camera inasmuch as a long development time 
produces graininess, which becomes conspicuous on enlarge- 

The correct exposure may be determined by means of an 
exposure meter and the tables and scales supplied with it. 
The photronic*-cell type of exposure meter is especially 
useful, but it has the undesirable feature that its response 
represents an average of the illumination received over a 
large solid angle. When the exposure is to be adjusted 
to the illumination of some small object which is much 
brighter or darker than its surroundings, it is necessary to 
put the photronic meter close enough to the object so 
that it substantially fills the field of view of the meter. 

* Registered trademark designating photoelectric cells and photoelectric de- 
vices manufactured by the Weston Electrical Instrument Corporation. 

Chap. XI] 



Tables and calculators like the Wellcome calculator are 
useful for roughly estimating the exposure. 12 To increase 
the chance of getting correct exposure, one should take 
auxiliary pictures at one half and twice the exposure pre- 
dicted by the calculator. Such a practice is especially 
feasible with miniature cameras, for which the cost of film 
is small. 

The speed of a given photographic material varies with 
the color characteristics of the illumination, depending upon 
whether it is sunlight, arc light, or light from a tungsten 
lamp. The exposure can be accurately estimated only after 
taking account of a color factor as well as the intensity of 
illumination. The success of an estimate depends largely 
upon the experience of the photographer. 

When unfamiliar conditions of illumination are encoun- 
tered, the correct exposure is usually determined by trial. A 


Exposure Characteristics of Different Photographic Emulsions for 
Various Light Sources 


Crater of 




4000 °K. 






Super-speed motion- 
picture film 

Par-speed motion- 
picture film 

Motion-picture posi- 
tive film 



















Bromide paper 


Hardy, Arthur C, J. O. S. A., 14, 515 (1927). 

Exposure required to give a perceptible image on various photographic 
materials. The values given in the table are B/<r X 10~ 6 , where B is the 
intrinsic illumination in candles per square centimeter and <r is the exposure to 
give a perceptible deposit in lumen seconds per square centimeter. 

12 This calculator may be obtained from Burroughs Wellcome and Company, 
9 East 41st Street, New York City. 


succession of trial plates is taken, the exposure of each vary- 
ing by a factor of 2, 3, or 5. 

Table VI is useful for estimating exposure. This table 
gives the ratio of the intrinsic illumination B to the exposure 
a required for the various light sources to yield a perceptible 
deposit on the photographic material indicated in the left- 
hand column of the table. 13 B is in candles per square centi- 
meter and <t in lumen seconds per square centimeter, so that 
the ratio B/cr has the dimensions 1/ (solid angle X seconds). 
If the product of the quantity B/cr multiplied by the solid 
angle of the illumination cone on the film and as well by the 
exposure time is equal to or greater than unity, a perceptible 
image will result. 

The values of B/cr are useful for designing recording in- 
struments to determine the maximum velocity v at which 
the recording spot can traverse the film and yet produce a 
readable trace. Let us consider a recording system, say one 
to record galvanometer deflections. First, we must deter- 
mine the exposure time and solid angle of illumination. If 
the width of the spot is d, then d/v gives the exposure time 
for those areas of the photographic material which have been 
traversed by the spot. The solid angle of the illumination is 
determined as follows: 

Case I. For the case in which a galvanometer mirror of 
area A acts as field stop and the light is focused on the film 
by a spherical lens immediately in front of the galvanometer, 
the solid angle is A//i 2 , when /i is the distance of the film 
from the spherical lens. 

Case II. For the case in which an astigmatic optical 
system is used, wherein the width of the galvanometer 
mirror, w, determines the lateral field stop, and the stop on 
the cylindrical mirror is h, the solid angle is 

2 v-. 
/i /. 

Here, /i is the distance of the film from a spherical (or 

13 Hardy, A. C, J. 0. S. A., U, 505 (1927). 


cylindrical) lens in front of the galvanometer mirror, and f 2 
is the distance of the film from the cylindrical lens near 
the film. (See Fig. 47, Chapter IX.) 

Accordingly, for Case I the condition for obtaining a 

record is 

B A_ d 

and for Case II, the condition is 

B w h d ^ 
a /i /a \ V ~~ 

Development. The aim of development is to render in 
photographic blackening the variations in illumination regis- 
tered as the latent image in the photographic emulsion. 
Sometimes, as in a snapshot, one may wish the rendition to 
be "normal," so that the positive print will seem to represent 
faithfully the original scene. On the other hand, one may 
wish to repress or enhance contrast by changes in the devel- 
opment procedure. 

The development process is not completely understood. 
Grains of silver bromide which have been exposed to light 
are reduced to metallic silver by the developer, while those 
not so exposed are not easily reduced. The developer will, 
however, finally reduce unexposed grains. The reduction of 
unexposed grains produces a general fogging of the plate, 
called chemical fog. 

The progress of development with time is illustrated in 
Fig. 11. The various H and D curves shown here represent 
a series of exposures, on five different plates. Each of the 
plates was developed for a different length of time, namely, 
2, 4, 8, 16, and 32 minutes. It is characteristic of develop- 
ment that the contrast increases with the time of develop- 
ment. At first the increase is rapid; for example, in 2 
minutes the contrast of the material represented in Fig. 11 
increased from 0.32 at t = 2 minutes to 0.58 at t = 4 min- 
utes, an increase of 0.26. Later, the rate of increase falls 



[Chap. XI 

off; for example, in 16 minutes the contrast increased by 
only 0.3 from 1.4 at t = 16 minutes to 1.7 at t = 32 minutes. 
The contrast approaches a limit 7^ for very long develop- 
ment. We have referred to this quantity before and tabu- 
lated it for representative photographic materials. (See 
Table III.) An interesting geometrical feature of the H and 

0.001 0.01 0.1 1 

exposure (lumeny meter*- seconds) 
Fig. 11. After Hardy and Perrin. (See footnote 1.) 

D curves shown in Fig. 11 is that their extrapolated straight 
segments have a common intersection point with the line 
A = 0. 

Developers are composed usually of four constituents: 
(1) The reducing agent — metol (elon), pyrogallol, glycin, 
amidol, or hydroquinone; (2) the alkali accelerator — 
caustic soda, sodium carbonate, or borax; (3) the preserva- 
tive — usually sodium sulphite or bisulphite; (4) the restrain- 
ing agent — potassium bromide. 

Most developing solutions will not develop at all unless 
they are alkaline, and all of them act more rapidly in propor- 
tion to the concentration of the alkaline accelerator. 

Oxygen dissolved by alkaline solutions may oxidize the 
reducing agent. In the case of pyrogallol this oxidation is 
particularly objectionable, for it yields a yellow-colored 
product which stains the emulsion. Oxidation can be 
avoided if the preservative sodium sulphite is added to the 


developer. This substance rather than the pyrogallol reacts 
with the dissolved oxygen, forming sodium sulphate. 

The characteristic effect of the restraining agent, potas- 
sium bromide, is to depress the intersection point, referred to 
above, below the line A = and at the same time bromide 
suppresses development in the low exposure range. A small 
amount of bromide which is not enough to depress the inter- 
section point much will, however, exert a selective enough 
restraining effect on the development to inhibit the forma- 
tion of undesired chemical fog, without having any sensible 
effect on the development of the desired image. 

There are two general procedures for development: time 
and temperature, or tank development; and factorial, or 
tray development. By the first method, the film is im- 
mersed in a tank of developer for a prescribed length of time. 
This time is determined by the nature of the film, by the 
degree of contrast desired, and also by the type, concen- 
tration, age, and temperature of the developer. 

By the second method, the progress of the development 
may be watched. Panchromatic emulsions are desensitized 
to make this possible. The time at which the development 
is to be terminated is either determined by inspection or 
calculated from the time required for the image first to 

Time and temperature development. The rate of develop- 
ment, as in any chemical reaction, increases rapidly with 
temperature. It is a practical rule in chemistry (although 
not a very rigid one) that a reaction rate increases by a factor 
of 2 for each 10° rise in temperature. This rule applies to 
development; for example, with pyro-soda developer this 
factor is 1.5. Time-temperature tables are available for the 
various developers. However, to get proper contrast and to 
prevent excessive graininess and fogging, it is advisable to 
carry on development at the temperature specified by the 
formula. For developers containing both metol and hydro- 
quinone it is particularly important to develop at the tem- 
perature specified in the formula in order to obtain the 



[Chap. XI 

proper proportionate effect of each ^-iujer. Metol has a 
low temperature coefficient, while hydroquinone has a high 
one. As a matter of fact, some operators take advantage of 
this difference and use hydroquinone-metol developers warm 
to obtain one grade of contrast and cold to obtain another. 
This procedure is not recommended, but it does illustrate 
the point in question. 

The common developers may be made up from the stable 
stock solutions listed in Table VII. These solutions are 
combined in the proportions given in Table VIII at the time 
they are needed, a procedure which is at the same time both 
economical and convenient. 14 


Stock Solutions for Making Various Developing Baths 



Metol or elon 2.5 g 

Anhydrous Na 2 S0 3 18 g 

Water to make 200 cc 

Hydroquinone 6.7 g 

Anhydrous Na 2 S0 3 12 g 

Water to make 200 cc 



Anhydrous Na 2 C0 3 .400 g 

Water to make 2 liters 

KBr 10 g 

Water to make . . . 100 cc 



Anhydrous Na 2 S0 3 400 g 

Water to make 2 liters 

Borax 10 g 

Water to make 250 cc 

Solutions C, D, E, and F are very stable and may be prepared and kept in. 
large stock bottles. Solutions A and B are less stable, and therefore should be 
stocked in small bottles to avoid an excessive amount of air over the solutions. 
Use distilled water for making the solutions. Dissolve chemicals for Solutions 
A and B in the order listed. 

14 1 am indebted to Dr. John McMorris for suggesting these tables. 

Chap. XIJ 




Proportions for Compounding Developing Baths from the Stock 
Solutions of Table VII 









Contrast Plate: D-28 













Process tank or tray: D-ll 

Fine grain: D-76 


D-72 (Chloride paper) 

D-72 (Bromide paper) 

The quantities are parts, by volume. The developer is compounded in the 
order A to E in the expressed proportions and then diluted to 100 parts, by 
volume. For the first three formulas, develop at 65 °F. (18°C). For papers, 
develop at 70 °F. (21 °C). See Elementary Photographic Chemistry for further 

Developers such as Sease 3 give fine grain, but they re- 
quire extra exposure. So-called compromise developers, 
such as Edwal 12, yield fine grain, and at the same time they 
do not require excessive exposure. Compromise developers 
were first made for use with miniature cameras. Recently, 
however, they have had some application in astronomy. 15 
Formulas for both the Sease 3 and Edwal 12 developers are 
given below. 16 

Sease 3 
(for twice normal exposure) 

Sodium sulphite (anhydrous) 90 g 

Paraphenylene diamine 10 g 

Glycin 6 g 

Distilled water 1 liter 

Developing time: 30 minutes at 65 °F. (18 °C). 

With careful regulation of exposure and developing temperature, 
negatives are produced whi ,h can be enlarged to 50 diameters and 

15 Morgan, W. W., Astrophys. J., 83, 254 (1936). 

16 The Edwal Laboratories, 732 Federal Street, Chicago, Illinois. 



[Chap. XI 

Edwal 12 

(for normal exposures) 

Sodium sulphite (anhydrous) 90 g 

Paraphenylene diamine 10 g 

Glycin , 5 g 

Metol 6 g 

Water 1 liter 

Developing time: 12 to 18 minutes at 65° F. (18° C). 

This developer keeps well and will give a finer grain when it is a 
month or two old than when it is fresh. The fineness of the grain 
produced also increases after it has been used the first time. 

For obtaining fine grain and absence of reticulation, it has 
been recommended that the temperatures of the developer, 
rinse water, fixer, and final wash water all be equal to within 
± 2°F. 

Over-all density is not important for films or plates that 
are to be printed. The recommended procedure is to dis- 
regard density in development. One develops for the desired 
contrast rather than for a specified average density. 

close-fitting cover 
made of paraffin 

battery jar 

Austin battery jav 
(about 6x6x4") 


short stop 

Fig. 12. 

battery jar 
with paraffin 



Fig. 12 shows equipment for developing. The paraffin 
covers shown protect the solutions from oxidation and 
evaporation when they are not in use. 

Tray development. When the emulsion is immersed in the 
developer, the time for the first appearance of the image is 

Chap. XI] 



proportional to the time for its full development. Accord- 
ingly, the time for first appearance of the image may be used 
to determine the proper time of development. This is help- 
ful when the concentrations of the chemicals in the develop- 
ing bath are different from those specified in the formula. 
In such a case one would not otherwise know how long to 
develop. The ratio of the time required for proper develop- 
ment to the time required for the image to appear is a char- 
acteristic of the developer and is called the Watkins factor. 
This factor varies with temperature for two-reducer devel- 
opers like metol-hydroquinone; but for others it can be used 
to correct for high or low temperature as well as for devia- 
tions in chemical composition of the developer. Watkins 
factors for several developers are given in Table IX. 

Watkins Factors 


Watkins Factor 









The Wellcome Handbook. 

When one is working for high contrast under conditions 
which allow observation of the progress of development, the 
rule is to develop in D-28 or D-ll until the image appears on 
the back of the emulsion. This rule applies particularly to 
the production of good lantern-slide and spectrum plates. 

Aerial fog is often encountered in tray development. It is 
caused when the plates, wet with developer, are exposed to 
the air. " Seeding' ' the developing solution with 5 per cent 


of old developer, which is rich in bromide, has a negative 
catalytic action on this fogging. The metals tin, copper, 
zinc, and their alloys often produce fog and stain on the 
emulsions when they come in contact with the developers in 
which the emulsions are treated. For this reason, develop- 
ment in brass and soldered metal trays is to be avoided. 

Plates for photometric work should be developed in a tray 
in a deep solution and brushed continuously with a camel's- 
hair brush. This accelerates convection near the emulsion 
and results in more even development. Regions in which 
the film has a strong exposure give off bromide during devel- 
opment, and unless this is removed by brushing, it restrains 
further development both at the place it is generated and at 
adjacent regions in the emulsion. 

Plates and films should be held by their edges, ends, and 
corners, and the fingers should not be allowed to come in 
contact with the important areas of the emulsion. In some 
cases the use of rubber gloves is indicated. Important 
plates should be rinsed in water before development. 

Desensitizing. Panchromatic emulsions must ordinarily 
be developed in the dark. However, the Luppo-Cramer 
discovery of the desensitizing action of phenosafranine on 
the unexposed silver grains allows illumination of panchro- 
matic emulsions with a green safelight during the latter 
stages of development. The selective desensitizing action 
of this dye on unexposed silver bromide grains is quite mys- 
terious; the development proceeds after the light is turned 
on as it would in darkness. The plate is separately im- 
mersed in the desensitizing solution for 2 minutes in darkness 
before development (1 part of a stock solution, of | g pheno- 
safranine in 1 liter distilled water, to 10 parts water). Pina- 
kryptol green added to the developer may be employed as 
desensitizing agent. The stock solution is a 1:500 solution 
of the dye in distilled water — 2 to 3 cc of this solution is 
added to each 100 cc of developer. 17 

17 Neblette, C. B., Photography, page 298. New York: D. Van Nostrand 
Company, 1930. 


Fixing. After proper development and washing, it is 
often adequate to fix the plate or film by immersion in a 
20 per cent solution of sodium thiosulphate (hypo) crystals. 
The hypo dissolves out undeveloped silver bromide grains. 

The plate usually carries some developer into the fixing 
solution and if the fixing solution contains no preservative, 
the transposed developer will be gradually oxidized and form 
products which may stain the film. Accordingly, to avoid 
this effect, fixing solutions are usually compounded with the 
preservative sodium sulphite added to the hypo. A weak 
acid is also added to neutralize alkali brought in on the 
plates. The acid has the further function of stopping 
development. Fixing solutions may also contain hardeners, 
such as potassium or chrome alum. These hardeners "tan" 
the gelatin, prevent excessive swelling and softening of the 
gelatin, and make it less "soluble" in water. 

A short stop is used to stop development and to conserve 
the acid in the hypo. Plates are immersed in the short stop 
solution after the development and before fixing. 

Short Stop 

Water 1000 cc 

Acetic acid, 28 per cent 48 cc 

This solution is to be made up fresh each time it is used. It 
does its work in about 5 minutes. 

The Eastman bath F-5, made up according to the formula 
given below, is a good one, and it is recommended for fixing. 

Eastman Acid Hardening Fixing Bath F-5 

Water at 52°C 600 cc 

Hypo crystals 240 g 

Sodium sulphite (anhydrous) 15 g 

Acetic, 28 per cent 47 cc 

Boric acid crystals 7.5 g 

Potassium alum 15 g 

Water to make 1 liter 

The chemicals listed in the formula are to be dissolved in the 
order given. For best results one should use only fresh 
fixing solution. It is best to fix plates for twice the time 


required to clear the gelatin of unreduced silver salts. With 
acid in the fixing bath the room lights may be turned on a few 
seconds after the plates are immersed. 

Ordinary gelatin melts in water at about 40°C. Normal 
hardening increases the melting temperature to between 
55° and 77 °C. However, for extreme hardening, Formalin 
is used. In less than a minute a 5 per cent solution of this 
chemical renders the gelatin film insoluble, even in boiling 
water. Formalin does not work in acid solutions. The 
following formula is recommended by the Eastman Com- 

Formalin Hardener 

40 per cent formaldehyde solution 10 cc 

Sodium carbonate (anhydrous) 5 g 

Water to make 1000 cc 

Washing and drying. It is necessary to remove the fixing 
solution from the emulsion by thorough washing. Table X 
gives the minimum time required for washing various emul- 
sions when the surface of the emulsion is held under a tap so 
that it is continually in contact with fresh water. But when 
these materials are washed in a tray without agitation, the 
diffusion of the fixing chemicals through the emulsion pro- 
ceeds more slowly. As a working rule, the washing time 
should not be less than the time required to wash the emul- 
sion under the tap plus the time required to wash the tray. 
The latter time is determined by measuring the time required 

Washing Time for Various Emulsions 



Lantern-slide plates 


High-speed negative emulsions 


Chloride papers 

15 to 20 

Bromide papers 

20 to 60 

From Elementary Photographic Chemistry. 

Chap. XI] 



to clear the tray of a strongly colored ink solution. The 
rate of washing is roughly independent of temperature, and 
it is also unaffected by hardening if the hardened emulsion 
has not been dried. A final rinsing with distilled water is 
recommended. Fig. 13 illustrates methods for washing 
plates, films, and papers. 

strips of 
wire screen 
b&nt to 
the over- 
flow from 
one tray 
into the 




water overflows from 
the end com pel vtment 

.plates being washed 

guides to hold plates 

ridges to hold 
plates off the bottom 

waste water enters 

end compartment at the bottom waste 

Fig. 13. Washing of plates and papers. 



[Chap. XI 

Plates and films are usually dried in a current of dust-free 
air. The air can be slightly warmed, since the heat of 
evaporation will keep the emulsion cool. Apparatus suitable 
for this is illustrated in Fig. 14. Fig. 15 shows a drying 
cabinet for films. 

The dry emulsion on negatives is about 0.0005 of an inch 
thick and normally contains 8 to 16 per cent moisture. It 
is five to seven times as thick as this when it is wet. The 
swelling of gelatin is characterized by the fact that it is 
anisotropic, being perpendicular to the glass. Even when 

^.vent holes 

plate rack 
with sa.w 

hinged door 

oiled -cheesecloth 
dust catcher 

hair dryer 

Fig. 14. Drying cabinet. 

gelatin is free from the lateral restraint, offered by the glass 
support, the swelling is principally in one direction. How- 
ever, when the film dries unevenly, silver grains are shifted 
laterally. The edge of a plate dries first, and the shifts 
produced are appreciable as far back from the edge as 1 cm. 
Drops of water (tears) or the excessive water held around a 
dust particle will produce lateral shifts of the emulsion. To 
avoid these shifts, the emulsion may be wiped with a damp 
cotton pad, a chamois skin, or better yet, a cellulose sponge. 
An automobile windshield wiper can be used to remove 
tears from plates and films. 
f The plate can be dried in 80 per cent alcohol when maxi- 

Chap. XI] 



mum accuracy to the edge of the plate is desired and also 
when distortions such as those caused by dust specks and 
tears are to be avoided. The alcohol is then evaporated 
from the emulsion by a current of moist air. If the alcohol 
or the air is too dry, the water in the gelatin will be reduced 
below the normal amount, and this subnormal water content 


holes covered 
with cheese- 

unpa'tnted clothes- 
pins, drille/d and 
strung" on a stiff wire 

paper mat 


films to be 

Fig. 15. Drying cabinet for films. 

turns the gelatin white. The water content of the gelatin 
can be raised to remove the milky appearance by breathing 
on the emulsion once or twice. 

Printing and enlarging papers. Negatives are printed on 
so-called printing-out papers, gaslight papers, or bromide 
papers. Printing-out papers are exposed to light through 
the negative until the image develops. The chloride (gas- 
light) papers are exposed and then developed with chemical 
developers. They are relatively insensitive and can be 
manipulated in a lighted room. Bromide papers must be 


managed in darkness or under a safelight, in the same 
manner as ordinary unsensitized photographic plates are 
managed. The chloride and bromide papers are the most 
important, and we will discuss them here. 

The reflectivity-density A for papers is defined as follows : 

A = logio-^- (6) 

R is the diffuse reflectivity of the paper. H and D curves 
for chloride contact and bromide enlarging papers, plotting 
A as ordinates against the logarithm of exposure as abscissa, 

chloride contact papers 

t = time, of 

10* iO* io 4 10* 

exposure (lumen>/meter 2 -seconds) 

Fig. 16. After Hardy and Perrin. (See footnote 1.) 

are illustrated in Figs. 16 and 17. It is to be noted that the 
contrast of these papers, or slope of the straight segment of 
the A-log 2 curve, does not change much with increasing 
development time. The contrast is more of an intrinsic 
property of the emulsion than it is with plates and films. 
Papers are obtainable commercially in various grades, so 
that if one paper does not give the desired contrast, another 
grade is used. 

The procedure for exposing and developing papers is 
different from that recommended for plates and films. 
Proper contrast is important with plates and films, while 
density is not so important. On the other hand, correct 
density is important for papers. The procedure for exposing 

Chap. XI] 



bromide paper, for example, is to adjust exposure so that 
development for 3 minutes yields proper density. 

Development for 3 minutes rather than the 2 minutes 
usually prescribed yields more contrast in the highlights, as a 
perusal of the curves of Fig. 17 will show. For the exposure 
range below 10 lumen seconds /m 2 , it will be noted that the 
inclination of the H and D curve to the abscissa is greater 
for the 3-minute development than it is for the 2-minute 

The proper exposure is determined with narrow test strips 
of paper. When the density is correct, the image shows a 


bromide enlarging p 


t = time of 

1 10 

10 2 

10 3 

exposure (lumens/meter*- seconds) 
Fig. 17. After Hardy and Perrin. (See footnote 1.) 

thin veil of blackening in the highlights, compared with the 
clear unexposed margins. A common fault in improperly 
printed negatives arises from overexposure and under- 
development. Never develop papers for less than 2 minutes. 

Table XI gives the relative speeds of different papers, so 
that once the exposure is determined for one grade, that for 
another may be calculated approximately. These values 
are not and cannot be precise because of unavoidable varia- 
tions in manufacture. Nevertheless, the table is of value for 
practical work. 

If, in development, a portion of the print fails to come up 
at a satisfactory rate, owing to an extra high local density 
of the negative, development may be accelerated locally by 


applying a piece of cotton wet with warm water or simply 
by warming the film with heat from the fingers, rubbed on 
the selected spot. 

Some of the organic developers, particularly amidol, are 
poisonous. The toxicity of developers varies for different 
people, but contact may be avoided if wooden tongs are used 
for manipulating prints in the developer. (See Fig. 12.) 
The print is moved around or the solution is stirred as 
development progresses. Care must be taken not to allow 
the wooden tongs to come in contact with the acid of the 
short stop or hypo of the fixing solution, since these chemicals 
would ruin the developer solution. 

Prints are washed conveniently in a cascade of flat pans as 
illustrated in Fig. 13. As the fixing of each print is com- 
pleted, it is put in the bottom pan. But before this is done, 
the print in the top pan is removed, and the print in the 
second pan is advanced to the top, and so forth, until the 

bottom pan is emptied to 
make ready for the one in the 
hypo bath. The prints, after 
washing, are laid emulsion 
side down on a cheesecloth 
tray to dry. 

As soon as the prints are 

dry, they are bent along their 

diagonals in the manner illus- 

Fig. is. trated in Fig. 18 to remove 

wrinkles. They are then ready 

for mounting. It may be desirable to flatten them out further 

by pressing them between the leaves of a book or magazine. 

Glossy prints are squeegeed on enameled plates, using a 
solution of paraffin or Ozokite in benzene to prevent the 
print from sticking. The solution should be rubbed dry 
and polished with cheesecloth before the print is squeegeed. 
Also, for drying glossy prints, chromium-plated brass sheets 
are now available which do not require paraffin or Ozokite, 
but only need to be washed and wet with water. 

Chap. XII 



condensing \ons focusing an image 
V of the light source on the enlarging lens 

■— .*_.. ^enlarging lens .. 

W \ negative > diaphragm 

lamp enlargement 

This, arrangement gives the greatest contrast. 

I j- flashed- of frosted-glass 
^ diffusing screen 
-^j*-.^---. m\ u* ''enlarging* lens 
ne^ * diaphragm ~ ' ^i 

lamp enlargement 

This arrangement gives less contract. 
Fig. 19. 

Fig. 19 illustrates two arrangements for enlarging. The 
so-called specular density is involved in the one shown above 
and the diffuse density in the lower one. The one above is 
generally used. The lower arrangement is used in por- 
traiture, or whenever contrast is to be avoided. 

Relative Exposures Required for Various Papers 





Bromide papers 

Chloride papers 




Photographs for publication which require strong contrast 
should be printed on glossy paper, which exhibits higher 
contrast than matte paper. 

Intensifying and reducing. The function of intensifying 
is to increase contrast, whereas the function of reducing is 
primarily to decrease the density. Reducing may, however, 
increase or decrease contrast or leave it unaltered, depending 


on the solutions used. The processes of intensifying and 
reducing will correct errors in the original development, but 
they will not satisfactorily correct errors of exposure. 

The chromium process of intensification will be treated 
here. It gives a permanent intensified image, in contrast 
to the mercury process, which is not permanent. To apply 
the chromium process, the gelatin should be first bathed in a 
hardening solution. After this, the emulsion is bleached 
clear in a mixture of 1 part of the following stock bleaching 
solution with 10 parts water. 

Bleaching Solution for Chromium Intensifier 

Potassium bichromate 90 g 

Concentrated hydrochloric acid 64 cc 

Water to make 1000 cc 

The bleached emulsion is then washed for 5 minutes and 
redeveloped, after which treatment it will exhibit greater 
contrast. It is then thoroughly fixed and washed. The 
process may be repeated to obtain even greater contrast. 
The intensification process does not need to be carried out 
in the dark. 

To obtain extreme contrast the intensified negative is 
printed on a transparent film, which is then intensified and 
printed again, and so on, until the desired result is obtained. 

Intensification occurs with each photographic process, so 
that often in the second photographic process (printing) 
special care is necessary to prevent the development of ex- 
cessive contrast. It may be necessary to employ a reduction 
procedure. The various formulas and directions for reduc- 
ing, with or without change of contrast, are given in Elemen- 
tary Photographic Chemistry. 18 

Some applications of photography. Photographs may 
be taken in the infrared with a suitably sensitized emul- 
sion, an ordinary lens, and a suitable filter, such as the 88A 
filter shown in Fig. 7. Heated objects, even a hot electric 
iron, giving off entirely invisible radiation, can be used for 

18 This book is available from firms dealing in photographic supplies or 
directly from the Eastman Kodak Company, Rochester, New York. 


illumination. Hypersensitizing is particularly effective in 
infrared photography. 

R. W. Wood of Johns Hopkins University and W. H. 
Wright of Lick Observatory were the first to show some of 
the striking effects that could be obtained in infrared photog- 
raphy. The penetrating power of infrared photographs, 
giving clear pictures of objects many miles distant, even 
through light haze, accounts for the present wide application 
of infrared photography, especially for aerial mapping. 

Photographs taken with ultraviolet light have yielded 
equally striking results. They may be taken through any 
of the filters that are opaque for the visible spectrum, such as 
bromine vapor, nickel sulphate solutions, thin silver films, or 
alkali-metal films. The image in ultraviolet light may be 
formed by a thin quartz meniscus lens or a quartz fluorite 
achromat. Ultraviolet photographs taken out of doors are 
free of shadows. 

From photographs of the moon taken through a silver 
film, as well as photographs taken with a bromine-vapor 
filter, R. W. Wood ascertained that a spot close to the crater 
Aristarchus was apparently covered with a layer of sulphur. 
Sulphur exhibits a high reflectivity in the visible spectrum 
and a very low reflectivity in the ultraviolet. The crater 
appeared black in Wood's ultraviolet pictures, in strong 
contrast with its surroundings. No other substance on the 
moon has been identified by means of such convincing 
evidence. "> 

The camera can be used for making drawings. The 
object to be copied is photographed, and an enlargement is 
made of it on matte paper. This is traced as desired with 
India ink. The paper is then treated with a bleaching solu- 
tion and the photographic image is removed. India ink is 
waterproof and unaffected by the bleaching solution. 

The following hints may aid in obtaining good results in 
the photography of apparatus for publication. In general, 
the object should be photographed against a light back- 
ground, such as a white wall or sheet. Polished surfaces on 


the apparatus will produce halation. The places giving 
halation (they may be determined by taking a preliminary 
picture) must be covered with vaseline and whiting. Best 
results are obtained by side-lighting from an open window or 
other light source of extended area. A good artificial source 
is obtained with a battery of photoflood lamps placed in a 
box covered with either tracing cloth or some other trans- 
lucent diffusing material. To obtain depth of focus in 
photographing apparatus, a small aperture and long exposure 
should be used. The exposure and development should be 
managed to give contrast and to show detail in the shadows. 

Miniature cameras are suited to making quantities of pic- 
tures under similar conditions (35 may be taken at each 
loading). They can be used for copying reference journal 
articles and even books. The negatives themselves may be 
used as the record, in which case they may be read with the 
film turned upside down in an improvised viewing box 
equipped with a strong reading lens, or they may be pro- 
jected on a screen. Also, the negative can be printed on 
motion-picture positive film. This does not require an 
excessive amount of time, since all the pictures have the 
same density, and they are processed in lots of 35 at a time. 
The miniature camera and projector is also useful for lec- 
tures. Graphs and illustrations can be printed on a single 
film in the order required by the lecture. 

The photographing of line drawings in pencil is difficult. 
A process emulsion should be used, and the illumination 
should strike the paper at oblique incidence to avoid specular 
reflection from the graphite pencil marks into the camera 
lens. The exposure and development are managed in such a 
way as to give maximum contrast. 

The details of the application of photography to spectros- 
copy and astronomy can only be touched upon here. In 
general, process plates are recommended for photographing 
spectra. They are best developed with D-28 or D-ll 
developer until the developed image shows up through the 
glass on the back of the emulsion. 


The region in the A-logS curve lying well within the 
latitude range of the emulsion is most suitable for photo- 
graphing spectrum lines. The most desirable density is 
about 0.5. The optimum light intensity for observation of 
the lines on a spectrum plate has a brightness value of about 
100 candles/m 2 . For much brighter or weaker lights the 
contrast sensitivity of the eye decreases. " Mixing up the 
grains" along the length of a weak spectrum line will ma- 
terially increase its conspicuousness. 

Photographic photometry. A spectrogram or stellar 
photograph usually serves its purpose if the intensity of the 
light is indicated approximately, provided the angular dis- 
tribution of the light is recorded with precision. The pecu- 
liar suitability of photography for observations of this kind 
is illustrated by the reference given in a footnote below. 19 
Photography may, however, be used to measure the intensity 
of light quantitatively. 

In photographic photometry, the plate is ordinarily used as 
an indicator to show, by equal densities, when the exposures 
of two areas are equal, one of the areas being produced by a 
source of known intensity, and the other by the radiation to 
be measured. It is important that the two exposures be 
made under equivalent conditions, that is, equal illumination 
and time of exposure or equal intervals of illumination if the 
light is intermittent. Also, it is important to take precau- 
tions against errors arising from variations in sensitivity over 
the surface of the plate. 

19 This property, which everyone now takes for granted, was explained by 
Fox Talbot in the early days of photography as follows: 

"Groups of figures take no longer to obtain than single figures. . . . the 
camera depicts them all at once, however numerous they be. . . . 

"It is so natural to associate the idea of labour with great complexity and 
elaborate detail of execution that one is more struck at seeing the thousand 
florets of an Agrostic depicted with all its capillary branchlets (and so accu- 
rately that none of all this multitude shall want its little bivalve calyx, requiring 
to be examined through a lens) than one is by the picture of the large and 
simple leaf of an oak or chestnut. But in truth the difficulty is in both cases 
the same. The one takes no more time to execute than the other for the object 
which would take the most skillful artist days or weeks to trace or to copy is 
effected by the boundless powers of natural chemistry in the space of a few 



[Chap. XI 

We will describe two methods of spectral photometry here. 
There are many possible sources of error in photometric 
work, and the treatment given here should be supplemented 
by reference to articles on this subject listed below. 20 

bar rnagfnet to be 
rotated by a 
hand magnet 

wedge being 
made (strip 
of glass clipped 
to the column) 

razor blade 
carriage with a 
shield to cover 
unexposed portion 
of glass | scr< 


to prevent the 
carriage from 
rotating, it 
must have an 
jirrn bearing 
the back 
of the 

Fig. 20. Making of a step-weakener. 

By the first procedure which we will discuss, a so-called 
step-weakener is mounted directly in front of the entrance 
slit of the spectrograph. For the first exposure, the slit is 

20 Abney, W. de W., "On the Variation in Gradation of a Developed Photo- 
graphic Image When Impressed by Monochromatic Light of Different Wave- 
lengths," Roy. Soc, Proc, 68, 300 (1901). 

Harrison, G. R., "Instruments and Methods Used for Measuring Spectral 
Light Intensities by Photography," J. 0. S. A., 19, 267 (1929). 

Jones, L. A., "Photographic Spectrophotometry in the Ultraviolet Region," 
National Research Council, Bull , No. 61, 109 (1927). 

Jones, L. A., and Sandvik, O., "Spectral Distribution of Sensitivity of 
Photographic Materials," /. O. S. A., 12, 401 (1926). 

Chap. XI] 



illuminated through the step-weakener with the light to be 
measured. For the comparison exposure the slit is then 
illuminated an equal time by the light of known intensity 
and known spectral distribution, this exposure being made 
adjacent to the first exposure. 

The step-weakener may be a glass or quartz plate coated 
with strips of platinum (or some other metal) of increasing 
opacity, or it may be a series of neutral filters of gelatin. 
The illumination along the slit, which would otherwise be 
uniform, is attenuated by the step-weakener in varying 
amounts, usually in geometrical proportion. 

detail of 
the raster 


Rowland circle,. K 



tangent to Rowland 
circle at intersection 
vVwith optic axis of 
\ v «rrcitin« 

concave grating 


Fig. 21. 


i << 


condensing lens 

The step-weakener is difficult to make and calibrate. 
Fig. 20 illustrates a procedure for making one by evaporation. 

From the calibration of the step-weakener at the wave 
length in question one can determine the relative light inten- 
sities of the different steps. To obtain the calibration, the 
densities of the various strips of the step-weakener are 
measured on a microphotometer. 

A photograph of the raster shown in Fig. 21 or a raster cut 
out of thin sheet metal can be used instead of a step-weak- 
ener. The use of a raster avoids the necessity for calibra- 
tion. In a uniform beam of parallel light the quantity of 
light transmitted by each element of the raster is deter- 



[Chap. XI 

mined by its area. The use of a raster requires an astig- 
matic optical system. The raster is ordinarily mounted so 
that, in a vertical azimuth its image is focused on the slit, 
whereas in a horizontal azimuth the parallel light transmitted 
by it is focused on the slit. 

Fig. 21 shows the ingenious way in which Frerichs has 
used a raster, taking advantage of the natural astigmatism 
of a Rowland grating setup. 21 In this setup, adjacent areas 
of the spectral image vary in intensity in the ratio 16:1 :2:4:- 
8:16. Fig. 22 illustrates the method of determining the 
relative intensities of two or more lines. The lines must be 
near each other in the spectrum, so that the differences in 


1 1 




i 1 



I 1 

*> * 


1 1 



1 1 



B C 

of lines A.B.C 

typical spectral lines 

1 2 4 3 16 

exposure (log scale) 

Fig. 22. After Frerichs. 21 

sensitivity of the plate for the wave lengths involved are 
negligible and so that the H and D curves for each wave 
length are the same. Three lines photographed through the 
raster by this setup are illustrated. The densities of the 
segments of the photographed lines are measured with a 
microphotometer, and plotted as shown. It is to be noted 
that the scale of abscissa is logarithmic, the interval in 
abscissa being equal for each step of the raster. Each 
spectral line determines an H and D curve, and the lateral 
displacements necessary to bring the three H and D curves 
into coincidence determine the relative intensities of the 
lines. By this procedure an accuracy of about 3 per cent 
can be expected. 

21 Frerichs, R., "Photographische Spektralphotometrie," Handbuch der 
Physik, Vol. 19, Chapter 23. Berlin: Julius Springer, 1928. 


Heat and High Temperature 

IN this chapter we will first consider some of the elemen- 
tary aspects of the theory of heat transfer. Following 
this we will deal with various techniques of obtaining high 
temperatures, of temperature control, and of temperature 

Heat conduction. The steady state. The rate $ (ex- 
pressed in calories per second) at which heat flows across an 
isothermal surface element of area A, in a homogeneous 
medium, is proportional to A, to the conductivity of the 
material, K, and to the temperature gradient dT/dx per- 
pendicular to the surface, thus : 

3> = — KA-j-~ calories/sec. (1) 

In the case of a rectangular parallelopiped with opposite 
ends maintained at the temperatures Ti and T 2 , Eq. 1, when 
integrated, becomes 

K A 

$ = _ — (T 2 - Ti) calories/sec, (2) 


in which A is the cross-section area of the parallelopiped 
perpendicular to the temperature gradient and x is the sepa- 
ration between the isothermal surfaces T± and T 2 . Here K 
is assumed to be constant in the temperature range between 
jPi and T 2 . Values of the heat conductivity for various 
materials are given in Table I. 



Thermal Conductivity, K, Diffusivity, h, and Relaxation Time, t,fob 

an Infinite Slab of 2 cm Thickness Cr = 1 cm) 


(at room temperature, unless 

otherwise specified) 



Brass (yellow)" 

Constantan (60 Cu 40 Ni) 6 




Wrought iron and mild steel" . . . 

Cast iron and carbon steel" 

Lead 6 


Monel metal"" 

Nichrome or Chromed 

Nickel* 1 



Tungsten: Room temperature* . 

1400°C. d 

2100°C. d 

Bonded silicon carbide, 25° to 

1000°C. e 

Graphite, 0° to 100° C. d 

Quartz glass: 0°C. d 

1000 o C. d 

Hard porcelain, 20° to 1000 °C. d . 

Fired natural soapstone* 


A1 2 3 * 

Sintered A1 2 3 , 900°C. d 
Mica* , 


Air, 0°C." 

Asbestos (loose)" 

Average firebrick, 0° to 800 °C.°. 
Concrete (stone), 20° to 1000°C. 

Cork (ground)" 

Paraffin" '. 


Pine wood: Across grain 

With grain 

























0.003 to 


0.003 to 


0.0015 to 


0.0016 to 



0.0008 to 












(cm 2 /si 
























0.0028 to 


0.00167 to 


0.0023 to 



0.00132 to 




























144 to 72 

240 to 120 

176 to 35 
306 to 174 





° Ingersoll, L. R., and Zobel, O. J., An Introduction to the Mathematical 
Theory of Heat Conduction. Boston: Ginn and Company, 1913. 

b Fowle, F. E., Smithsonian Physical Tables. Washington: The Smithsoni- 
an Institution, 1934. 

c King, W. J., Mechanical Engineering, 54, 275 (1932). 

d Espe, W., and Knoll, M., Werkstoffkunde der Hochvakuumiechnik. Julius 
Springer, 1932. 

• Hering, C, Am. L M. E., J., 29, 485 (1910). 



Shape factors. For many of the actual cases encountered, 
the geometry is not so simple as it is with the parallelopiped, 
and the integration of Eq. 1 may be quite difficult. Gener- 
ally, however, the integral may be expressed by an equation 
of the form 

$ = - KS(T 2 - Ti) calories/sec, (3) 

in which S, the so-called shape factor, depends upon the size 
and shape of the space between two isothermal surfaces 
maintained at temperatures Ti and T 2 . 

For a rectangular parallelopiped the shape factor, from 
Eq. 2, is A 

S = -■ (4) 

For two concentric cylindrical isothermal surfaces (long in 
comparison with their radii) of length Z, maintained at 
temperatures T x and T 2 , respectively, the value of S in terms 
of I and their radii n and r 2 is 

3 = 2.73Z (5) 

logi r 2 /ri 

For two concentric spherical isothermal surfaces, one of 
radius n at temperature Ti and the other of radius r 2 at 
temperature T 2 , the shape factor is 

S-.T^W (6) 

A heat problem which often arises in the laboratory 
requires the estimation of the heat loss of an electric furnace. 
The inner furnace wall, which is approximately at uniform 
temperature, is taken as one isothermal surface, and the 
outer surface of the furnace, at somewhat above room 
temperature, is taken as the other isothermal surface. Lang- 
muir, Adams, and Meikle have given shape factors for 
several special cases which may be applied to problems of 
this type. 1 However, to make an estimate of heat loss of a 

1 Langmuir, L, Adams, E. Q., and Meikle, G. S., Amer. Electrochem. Soc^ 
Trans., 24, 53 (1913). 






see fig. 17 
Fig. I. 

cylindrical furnace, if the inner furnace tube is long and 
surrounded by a layer of insulating material as shown in 
Fig. 1, we may apply Eq. 5. Or, for the case shown in 

p^ ^ Fig. 2, we may apply Eq. 6, 

1 taking r\ and r 2 as the dimen- 
sions of the approximating 
spherical surfaces, indicated 
in the figure by dotted lines. 
These estimates are not ex- 
pected to be precise, but they 
are usually accurate enough to 
settle the questions which arise 
when one designs a furnace. 
The shape factor can also be determined experimentally, 
using the similarity between the law for the flow of heat, 
Eq. 1, and Ohm's law. The experimental determination of 
S is accomplished by measuring the electrical shape factor, 
S', for wooden models that simulate the inner and outer 
isothermal temperature surfaces of the heat problem in 
question. These model surfaces are coated with copper foil 
and serve as electrodes. The region between these surfaces 
is filled with a saturated solution of copper sulphate with 
| per cent (by volume) sulphuric acid added. The con- 
ductance of this solution is deter- 
mined by applying alternating volt- 
age to the copper electrodes. Alternat- 
ing current is used to prevent polariza- 
tion at the electrodes. The equation 
giving the electrical shape factor is 

i = - K'S'V, (7) 

see Fig. 19 
Fig. 2. 

V and i being the measured voltage 

drop and current. K', the electrical 

conductivity of the solution, may be 

determined by a separate experiment, using a box of cross 

section A' and length x' with copper end plates. For this 

box the shape factor is A f /x r (as in Eq. 4). To transform 


S' to S, divide S' by the scale factor to which the model 
was constructed. For example, if the model was made to 
half scale, S = 2S'. 

Heat conduction. The nonsteady state. The thermal 
behavior of a homogeneous body is described in a Cartesian 
system of coordinates by the following fundamental differ- 
ential equation: 

8T KfFT ,&T. 8*T\ 

Here t is the time, T is the temperature of a point in the 
body represented by the coordinates x, y, and z, dT '/dt is the 
rate at which this temperature changes, and K, p, and c 
represent physical quantities for the material of which the 
body is composed, namely, the heat conductivity, density, 
and specific heat. The combination of these constants in 
the form h 2 = Kpc is convenient, h is called the thermal 
diffusivity of the material. 

In one dimension, Eq. 8 takes the form 

dT ^Kd?T 

dt P c dx 2 ' W 

If dT /dt equals zero, and if we integrate d 2 T/dx 2 once, we 
get the equation which represents the steady-state problem: 

£'->■ (l0) 

From physical considerations, the integration constant M 
is seen to have the meaning 

M = - £j> and $ = - KA^-> (11) 

KA dx K ' 

which is the same as Eq. 1. 

A more general form of Eq. 8 includes an added term to 
take account of energy transformations associated with a 
change of state, and so forth, which will not be considered 

There are infinitely many solutions to the fundamental 
differential equation, Eq. 8. Those which are appropriate 


for a given problem usually comprise an infinite series, the 
sum of which conforms to the requirements of the geometry 
of the body, and to the so-called boundary conditions set 
forth in the problem. The mathematical procedures in- 
volved in getting the series required for a particular problem 
were originally developed by Fourier over a hundred years 
ago; and these procedures have been extended by other 
mathematicians to include a great variety of more or less 
complicated cases. 2 Here, without taking up the mathe- 
matical procedures involved, we will discuss the results of 
their application to some typical heat problems. 3 

The infinite slab. First, let us determine the temperature 
at various points in a plane-parallel slab which, to start with, 
is at a uniform temperature T . We will find the tempera- 
ture at various places in the slab as a function of the time 
which elapses after the slab has been immersed into an 
environment maintained at a fixed temperature T±. We 
will assume that T 1 is lower than To. (The changes required 
to apply the results so obtained for the opposite case, in 
which Ti is higher than T Q , are obvious.) Practically, if the 
extension of the slab is great compared to its thickness, this 
becomes a one-dimensional problem, and to describe it we 
will take a Cartesian coordinate system which is oriented so 
that the faces of the slab coincide with the planes x = + x 
and — x . 

The solution of Eq. 8, which we want, is a series, the terms 
of which depend on both x and L The sum of the series 
yields a uniform temperature throughout the slab at t = 0; 
and also at all times it gives a temperature gradient at the 
surfaces which conforms to the requirements of Newton's 
law of cooling. 

Newton's law of cooling states that the heat lost per unit 

2 Carslaw, H. S., Introduction to the Mathematical Theory of the Conduction of 
Heat in Solids, Second Edition. London: The Macmillan Company, 1921. 

Ingersoll, L. R., and Zobel, O. J., The Mathematical Theory of Heat Conduc- 
tion, With Engineering and Geological Applications. Boston: Ginn and 
Company, 1913. 

3 1 am indebted to Dr. R. M. Langer for the treatment of nonsteady heat 
flow presented in this chapter. 


area of surface, W, by the slab to its environment, is propor- 
tional to the difference between the surface temperature T Xq 
and the temperature of the bulk of the medium in which it is 
immersed, 2\: 

W = N(T X0 - TO calories/sec./cm 2 . (12) 

W may be resolved into heat lost by radiation, TF rad ., and 
heat lost by convection, TT conv . The temperature gradient 
at the surface is determined by the value of W and the 
thermal conductivity of the material of which the slab is 

Stated algebraically, the boundary conditions which our 
solution of Eq. 8 must satisfy are 

at t = 0; T = To throughout the slab (13) 

and, for all values of t; 

at x = xo Tx = ~ TL^** ~ Tl) (14a) 

and also at 

dT N 
x = - xo °± = ±{T m - T x ). (14b) 

The solution of Eq. 8 which satisfies these conditions is 

... qo Sill —=j~ _ gnWhH 

T - Ti + -(T„ - ro£— 7 r re ix * cos M£, (16) 

\ a n ir J 

where the a„'s are roots of the transcendental equation 

2x Q N a n w 

^K = tan -T- (16) 

The values of a n may be determined graphically from the 
intersection points of the two functions of a n , 

y = 2^V' and!/ = C0t i- < 17) 

Before discussing various aspects of this solution, let us 
make the substitution, 

* = % ™ 


in the exponential terms, r is called the relaxation time. 
The reason for this will appear presently. 

At the beginning, that is, when t has small values com- 
pared with r, the accurate expression for T requires several 
terms of the series given by Eq. 15, in spite of the fact that 
the series is a rapidly converging one. However, soon after 
t = r, all the exponential terms become insignificant except 
the first one (n = 1). This is because ax is smaller than 
the other values of a n . Soon after t = t, Eq. 15 reduces to 

T = 7\ + -(To - 7\) 



sm— i 

2 — oi V ir x 

cos ai- 

flil + 

sin ai7r \ 2x 

• (19) 

The first factor in the brackets is a constant, the second deter- 
mines the decay of the temperature difference (T — Ti), 
and the third factor is the space distribution function for the 
temperature. The relaxation time is evidently the interval 
required for the temperature, initially uniform, to assume 
approximately the distribution given by the last factor in 
Eq. 19. 

The value of a\ for a body (with vertical sides) in air at 
room temperature is obtained from Eqs. 35., 48, 12, and 16: 

TFconv. = 1.3 X lO- 4 ^ - T xa ) calorie/sec./cm 2 . (35a) 

T^rad = 1.5 X 10- 4 (r - T X0 ) calorie/sec./cm 2 . (48a) 


N = 2.8 X 10~ 4 calorie/sec./cm 2 /°C. 

To illustrate how Eq. 19 may be applied, let us consider 
the case of a telescope mirror of 2 cm thickness which is to be 
tested by the Foucault knife-edge test. For a reliable test, 
if this mirror is brought from a room in which it is either 
warmer or cooler than the air of the testing room, it will be 
necessary to wait until the mirror has adjusted itself to the 
new temperature. If the glass is 15 cm or more in diameter 
and is exposed to the room air on both sides, we may regard 
it as an infinite slab and apply Eq. 19 to determine its thermal 


behavior. For the glass we may take K = .0024 and h 2 = 
.0057. This gives r = 71 seconds, and by means of Eq. 16, 
we get ax = 0.219. 

Substituting this value of a h Eq. 19 can be written in the 

logio( y* ^ y* ) = - 3 X 10- 4 * + 0.15 + log 10 (cos 0.34a;). (20) 

This solution is valid after more than 71 seconds have 
elapsed. To get the thermal behavior at the start, the 
logarithm of (Ti — T x )/(Ti — T ) can be plotted as ordinate 
against t as abscissa. The series of parallel straight lines 
obtained for t > 71 seconds are then extrapolated to the 
common point where the abscissa and the ordinate are equal 
to zero, bearing in mind that T x= ± Xo changes rapidly with 
time when t = and T x=0 changes very slowly. This method 
is not very precise, and a more exact solution is to be ob- 
tained from Eq. 15. This formula is rather difficult to 
manage, except in special cases. Two of these are treated 

Eq. 15 can be simplified for the extreme cases of relatively 
fast cooling, where iVaVK> > 1, and relatively slow cooling, 
where Nx / K < < 1. In the first case a n is approximately 
(2n + 1), sin(a n 7r/2) is (— l) n , and the expression for temper- 
ature simplifies to 

T = ^ + l^- T ^kfr * co S (2n + l)g.(21) 

For slow cooling, where Nx / K < < 1, the slab is practically 
isothermal, and the temperature is given by 


T = 7\ + (To - TO*** . (22) 

The solution of problems of this character will be useful 
to the experimenter when he encounters questions of design 
involving the accommodation of objects to changes of 


The application to optical testing has already been dis- 
cussed. In optical testing with the Foucault knife-edge 
test, lack of thermal equilibrium distorts the figure of an 
optical surface and gives rise to troublesome convection 

The relaxation time. The relaxation time for a cylinder is 
approximately half that for a slab, when 2x , the thickness of 
the slab, and 2r , the diameter of the cylinder, are equal. 
The relaxation time for a sphere or cube is approximately 
one quarter of that for a corresponding slab. In most of the 
nonsteady-state problems encountered, it is sufficient to know 
the relaxation time. The relaxation time can be interpreted 
as the time for a heat pulse to travel into the center of the 
slab, a distance x . The relaxation times are given in 
Table I for a slab thickness of 2 cm (x = 1 cm) for different 
materials. It must be remembered that for different values 
of x the time required for the heat to penetrate to the 
center of the slab is proportional to x 2 . 

The relaxation time for graphite, which is approximately 
the same as that for copper, is especially noteworthy. The 
extreme values for r given in Table I are about \ second and 
404 seconds for silver and paraffin respectively. 

Periodic temperatures. Let us consider a slab of thickness 
Xq having a harmonic surface temperature T\ = A cos ut. 
If r is the relaxation time for the slab, the interior tempera- 
ture is given by the expression 

4AwV£ (2n + l) 5 

(2n + 1)« 

7T 2d oV 

g> 2 t 2 C 

1 (2n+l) 4 

T = A cos ci>£ — 

- cos (at + (2w + l) 2 . I . . irx roo . 

sin g)^cos (2w -f 1)^— . (23) 

The exponential term can be neglected after the relaxation 
time, and the temperature is then given by the summation. 
Unless o)t > > 1, the convergence of the series is rapid 


r = 4|(l "',, cos ~ )cos ut 

enough to make the first term a good approximation for it: 


+ C(l+^V) ) Sln ^ C0S 2^)- (24) 
The product cor is the ratio of the relaxation time to the 

period of the impressed harmonic temperature multiplied by 
2w. If cor is small, the plate follows the impressed tempera- 
ture closely with an out-of -phase component, sin cot, propor- 
tional to cor, and the amplitude of the temperature fluctua- 
tion is proportional to cos (irx/2x ). 

When cor > > 1, the temperature near the surface is 
approximately the same as if the slab were infinitely thick, 
while the temperature in the center is practically- constant. 

The temperature at a distance x from the surface of an 
infinitely thick slab is given, after a long time, by the ex- 

T = Ae **<»s forf- J£?\ (25) 

where A cos cot represents the surface temperature. Thus, 
the amplitude decreases exponentially with depth according 

__ fix . . 

to the law e \2A, There is a time lag of -\/x 2 /2ooh 2 in its 
harmonic variation, relative to the phase of the surface 

The sphere. When a sphere or cylinder that is initially at 
a uniform temperature T Q is introduced into a medium at a 
lower temperature 2\, the equations similar to those for the 
slab are: 

For a sphere of radius r , 

. a n 7r . a n irr 

sm-4- «nw sin - 

O-rA -^ Sin a„7T J 

Tl +™g-(To - TO}} , -« 4ro ° — 21. (26) 

ttX *^ / a ra 7T 

1 r 

where a fl are the roots of 


tan M = 2^, (27) 

2 _ iVVo 



and after the relaxation time the term representing the 
temperature distribution (corresponding to cos {aiirx/2x^) in 
Eq. 15) is 

f • o. n Tcr\ 

When Nr /K ^ 1, the temperature is given approxi- 
mately by the expression 

• (2n + l)nr 
I '= y '+>»- r ^Z(feS)^"^~ -^L_.(28) 

When Nr /K > > 1, that is, for relatively fast cooling, 
the temperature is given approximately by 


X \n7rr 

sin ( 1 


In the case of slow cooling, in which Nr c /K < < 1, the 
temperature is sensibly the same throughout the sphere, and 
its change with time is given by the expression 


T = T x + (To - T x )e~ *'• . (30) 

The cylinder. For a cylinder of radius r , the temperature 
in terms of the well-tabulated Bessel functions J and J\ is 

t= r 1 +2(r -ri)?"/ 1 7#VV> / '" ^/m.A(S1) 

where the /x n 's are roots of the equation 

^t<A(mJ = ^-/oW- 

In limiting cases the /x ra disappear. For example, when 
Nr /K > > 1, that is, when we have fast cooling, the //„'s 


are close to the roots of Jo GO = 0, and the temperature is 
given approximately by the equation. 

T=T 1 + 2(T Q -T l ){^e Jo ( 2 . 4 [ X + _]-) 

i -C-4o(^^)] 2 

1.87 e 

+ etc. . (32) 

Temperatures at the center are obtained without the help 
of tables of Bessel functions because J(0) = 1. 

When Nr /K < < 1, with practically uniform tempera- 
ture throughout the cylinder, 

_ 2NhH 

T = T t + (To- rO e ** . (33) 

If Nn/K tt 1, the first term of Eq. 31 dominates after 
the relaxation time. The Bessel function can be expanded, 
and then the temperature is given by 


1 + ^F + 

Nr Q N 2 ri 

Heat transfer by free convection. Except for a few 
special cases, the estimation of heat loss by free convection 
is quite complicated or even impossible. The special cases 
which have been solved include plane surfaces and wires 
cooled by convection. The work on this subject up to 1933 
has been summarized by W. J. King. 4 Among the various 
methods for calculating convection losses, that of Langmuir 
is the simplest. 5 His method applies when the surfaces are 
small, such as those encountered in the laboratory. As it 

4 King, W. J., Mechanical Engineering, 54, 190, 275, 347, 410, 492, 560 (1932). 

5 Langmuir, L, Amer. Electrochem. Soc, Trans., 23, 299 (1913); Phys. Rev., 
34, 401 (1912). 

Rice, C. W., International Critical Tables, 5, 234. New York: McGraw- 
Hill Book Company, 1929. 



applies to a vertical surface, his method consists of calculat- 
ing the heat conduction through a postulated stagnant air 
film of 0.45 cm thickness, thus : 


Ti) calories/sec. /cm 2 . 


Here K is the thermal conductivity for air, Ti is the absolute 
temperature of the vertical surface, and T 2 is the ambient 
temperature. A more complete theory shows that W is 
proportional to (T 2 — T r i) 5 A and to the fourth root of the 
height of the vertical surface. K, in Eq. 35, is not in- 
dependent of temperature, and, except for small temperature 
drops, the heat transfer is given by the expression 

— 1 r Tl 
W = — - I 

0.45J T , 

KdT = - 


calories/sec. /cm 2 . (36) 

Values of <j> for air are given in Table II to facilitate calcula- 
tion. These values are defined by the expression 

T KdT. (37) 


Values of <f> for Air 


Calories /sec. /cm 

































Langmuir found that heat losses by free convection from a 
horizontal surface facing upward are 10 per cent greater than 
fchey are from a vertical surface, and they are 50 per cent less 
from a surface facing downward than they are from a vertical 

The procedure for calculating the convection losses from 
wires is also treated by Langmuir. 

Heat transfer by radiation. The energy emitted by a sur- 
face of area A radiating the heat spectrum between the wave 
lengths X and X + d\ is 

$dx = ire\J\d\ calories/sec. (38) 

This represents the summation of energy in respect to the 
solid angle over the hemisphere (angle 2w steradians). Here 
eX is the emissivity of the surface. This is the ratio of the 
emission of the surface to that which would obtain for a 
" black body" at the same temperature. JX is the energy 
radiated per unit solid angle by a black body of the same 
area at wave length X for a unit wave-length range, d\ = 1 cm. 

So-called black-body radiation is defined as the thermal 
radiation coming from the surface of a body which is in 
temperature equilibrium with all of its surroundings. For 
example, the inner surface of a cavity in an opaque material 
at a uniform temperature emits black-body radiation. In 
fact, black-body radiation is obtained experimentally from 
just such a cavity. The wall of the cavity is pierced to 
form a small aperture to serve as an outlet for the radiation, 
the hole being small enough not to disturb the equilibrium 
perceptibly. The name black-body radiation comes indirectly 
from KirchofTs law, which states that the emission and 
absorption coefficients of any body are equal. A black body 
with an absorption coefficient of unity, ak= 1, therefore, by 
KirchofTs law, has an emission coefficient of unity, cX = 1, 

The Planck expression for JX is a function of the wave 
length, X, and the absolute temperature, T. 


calories/sec. /cm /steradian. (39) 



wave length in ju 
Fig. 3. After Jean Lecomte. 

This formula describes the distribution of energy in the heat 
spectrum, and its plot against X at different temperatures is 
illustrated in Fig. 3. 

For XT = 0.3 this expression is approximated to within 
1 per cent by the so-called Wiens formula, 


J I = -r^e calories/sec. /cm/steradian. 


As XT becomes < < 0.3, J x becomes asymptotic to J x . 

For XT = 80 the expression is approximated to within 
1 per cent by the so-called Rayleigh-Jeans formula, 

J° x = — t~i T calories/sec. /cm/steradian. 


As XT becomes > > 80, Jx becomes asymptotic to J x . 
The values of the constants d and c 2 , where X is expressed in 
centimeters, are Ci = 2.81 X 10~ 13 calorie /sec. /cm 2 /unit solid 
angle; and c 2 = 1.432 cm degrees. 

The total heat lost by a unit area of the surface of a " black 
body" is the quantity expressed by Eq. 38 integrated over all 
wave lengths. This gives Stefan's formula: 

$ = irfJ x d\ = AaT* calories/sec. (42) 

Most surfaces have a total emission which may be ex- 
pressed as 

$ = Ae T aT 4 calories/sec. (43) 



Here a has the value of 1.38 X 10 -12 calorie /sec. /cm 2 / 
degree 4 . 

The heat emitted by a flat surface of area A into a cone 
which is defined by a solid angle d£l is 

d$ = A cos 6 — ero-7 74 calories/sec. 



Here A cos 6 is the projected area of the source and d£l/ir is 
the fraction of the total heat emitted in the direction 6 
defined by the element dQ. 

e T is an emissivity averaged over all wave lengths, and it 
is ordinarily "constant" only for a small temperature range. 
For porous nonmetallic substances it is very nearly unity, 
regardless of the color of the material. Naturally, the visible 
color of a body does not determine its infrared "color." 
Some bodies, such as white lead, are almost completely black 
throughout the heat spectrum, while the reverse is true for 
other substances, notably soot and black paper, both of 
which are transparent for the long wave-length end of the 
heat spectrum. e T for aluminum paints, around room tem- 
perature, varies between 0.3 and 0.5. For nonmetallic pig- 
ment paints e T = 1. 

For clean metals, e T varies with the temperature in such 
a way that the total emissivity is conveniently represented 
by an expression of the form 

$ = AMT m calories/sec. (45) 

Here M and m are constants. Values of M and m for some 
common metals are given in Table III. 


Radiation Constants of Metals 


Temperature Range 




610 to 980 
640 to 1150 
463 to 1280 
700 to 1300 
325 to 1310 

7.16 X 10~ 14 
5.50 X 10~ 16 
2.39 X 10" 15 
7.65 X 10~ 18 
4.30 X 10- 13 









Smithsonian Tables, 1934, page 324. 


The heat transfer by radiation between two parallel black 
isothermal surfaces of area A at absolute temperatures Ti 
and jT 2 , which are separated by a small distance, is 

W = o-(T 2 4 - TV) calories/sec./cm 2 . (46) 

W = I.SSJ^Y - ft^Y) calories/sec./cm^. (47) 

vioooy " vioooy 

If the temperature difference, (T 2 — Ti), is small, this 
heat transfer may be expressed so: 

W = 5.5 X 10- 12 r 3 A!T calories/sec./cm 2 . (48) 

Thus, owing to the fact that the absolute temperature 
enters the expression to the third power, we see that the 
importance of radiation as an agency for heat transfer be- 
comes greater at higher temperatures, until finally, in com- 
parison, ordinary conduction becomes negligible. 

At extremely high temperatures, the action of an in- 
sulator is the same as the action of a radiation baffle or series 
of baffles. The effect of baffles can be illustrated by the 
example of two infinite plane-parallel black surfaces at tem- 
perature jTi and T 2 . If a thin black baffle is interposed 
between these two surfaces, the transfer b>y radiation is 
reduced to one-half its original value. Two baffles reduce it 
to one-third, three baffles to one-fourth, and so forth, and if, 
instead of black baffles, polished metal reflectors are used, 
the insulation effect is even greater. In high-temperature 
furnaces the furnace tube with its winding is frequently 
surrounded by a thin sheet of some metal like molybdenum 
to serve as a baffle to reflect back most of the radiant energy 
emitted by the tube and so to decrease the power required. 
Sometimes, too, a second refractory furnace tube may sur- 
round the first to act as an insulator. 

Low temperatures. Moderately low temperatures are 
obtained in the laboratory by immersion in baths of ice, salt 
and ice, dry ice, liquid air, and so forth. The various tem- 
peratures so attained are listed in Table IV. For obtaining 


extremely low temperatures, the methods required are very 

elaborate. 6 


Freezing Mixtures and Constant-Temperature Cooling Baths 


T (°C.) 

N a Cl, 33 parts, plus snow, 100 parts 

CaCl 2 + 6H 2 0, 100 parts, plus snow, 70 parts . 

Liquid nitrogen Boiling point 

Liquid oxygen Boiling point 

Solid CO2 Sublimation point 

Mercury Melting temperature . . . 




Methods of obtaining high temperatures. Flames. The 
use of flames affords the most simple and convenient means 
of obtaining high temperatures. 

checks blown out like a bellows 
to give a continuous blast 

brass blowpi 

r . 

charcoal or asbestos work pinned in place 
soldering block „ ys. with phonograph 
& / / ^ — needles 

adjustment sleeve 

alcohol lamp 

Fig. 4. 

Fig. 4 illustrates the use of a blowpipe with the alcohol 
lamp, showing how the cheeks are used as bellows to give 
continuous air pressure. 

6 Meissner, W., Handbuch der Physik, Vol. 11, Chapter 7. Berlin: Julius 
Springer, 1926. 


Fig. 5 illustrates the ordinary Bunsen burner. The 

Bunsen burner is simply a tube 
arranged with a fuel gas jet in 
the bottom and air ports in the 
sides near the bottom. It draws 
air in through the^e ports by in- 
jector action of the gas jet. The 
air drawn into the tube at the 
bottom is mixed with the fuel 
gas as it passes up through the 

Fig~iT Pilot attachment (b) is tube ^ and above the to P ° f the 
obtainable from the Forster Manu- tube this air reacts chemically 
factoring Company, 2916 Otis with the gas fuel to p roduce the 
Street, Berkeley, California, and & r 

attachment (c) is obtainable from name. 

the Central Scientific Company, The Bunsen burner draws 
Chicago, Illinois. 

only about hall as much oxygen 

through its ports as is required for burning the fuel. If 
more air were mixed with the 
gas, the velocity of propagation 
of the flame would be greater 
than the upward velocity of the 
gas in the tube, and the flame 
would " backfire." However, 
additional air required for com- 
bustion of the gas is supplied to 
the flame above the burner tube ; 
owing to the more abundant 
supply of air, at the edges of the 
flame the propagation velocity 
is greater than the upward ve- 
locity of the gas, so that the 
fire does not blow itself out. 

Natural gas, which contains 
less hydrogen than coal gas, has 
a much smaller flame velocity. 
(The heat of combustion 
and the chemical composition of some commercial fuel gases 

nickel .screen 



Commercial Fuel Gases 


B.t.u. per 
Cubic Foot 

Per Cent H 2 

Per Cent CO 

Per Cent 


Per Cent 


Coal gas 

Natural gas . . 
Bottled gas. . . 










Central Scientific Company Catalogue. 

are shown in Table V.) Accordingly, with natural gas there 
is a greater tendency of the burner to blow out than with coal 
gas. This has resulted in the invention of fixtures like those 
shown in Fig. 5(b) and (c), which serve to retard the upward 
velocity of a portion of 
the gas mixture. The 
flame formed by this 
slowed-up portion does 
not blow out, and it pre- 
vents the main flame 
from doing so. A small 
tube may be soldered to 
the burner as shown in 
Fig. 5 at (a) to act as a 
pilot as well as to pre- 
vent the flame from blow- 
ing out. 

The Meker burner is a 
Bunsen-type burner with 
the top of the burner tube 
flared out and fitted with 
a nickel grill. This is illus- 
trated in Fig. 6. The 
Meker burner can burn 

Meker burner #un$en burner 

Fig. 7. After F. Haber. 

coal gas with a higher air admixture than the Bunsen burner, 
because the grill, acting as a Davy lamp screen, prevents 



the flame from backfiring. The hot inner blue cone of the 
Bunsen flame is replaced here by an array of small cones, 

brass -tube 


brass -lobe / 
wrapped with cord^ 
for a handle 

length about 8 inches 

Fig. 8. After Ernst von Angerer 

tubes silver- 
soldered together 
*61 to*76 drill hole 
* ZZ drill hole 

one over each element of the grill. This array produces a 
flame which is both hotter and more uniform over an ex- 
tended area than the Bunsen flame. The temperature dis- 
tributions in the Bunsen and Meker flames, with coal gas 
fuel, are shown in Fig. 7. 

To obtain a higher temperature than either the Bunsen or 
Meker will yield, the fuel is burned with air or oxygen under 


Fig. o. The burner shown at the right, for natural-gas fuel, is obtainable from 
the Forster Manufacturing Company, 2916 Otis Street, Berkeley, California. 

pressure at the end of an orifice with burners such as the 
ones shown in Figs. 8 and 9. When natural gas is burned' 


with air, a special tube end is required. (See Fig. 9.) An- 
other method of burning gas to get a high temperature is to 

water faucet 

air intake (can 
be used for 


fuel >~^ A 
air or oxygen ^^ 

Fig. 10. 

Fig. 11. 

project a jet of air or oxygen through a gas flame as shown in 
Fig. 10. A simple method using a water aspirator for obtain- 
ing compressed air at moderate pressures is shown in Fig. 11. 

f ull j"**™""*"^ v 

Fig. 12. 

Extremely high temperatures are attained with oxy- 
hydrogen or oxyacetylene torches. Commercial torches like 
the one illustrated in Fig. 12 are recommended for these 
fuels. 7 These torches are equipped with a mixer, usually in 
the handle, to produce a 

homogeneous solution of ^^^^S^^^=^jy &dtiUfot ' 
the fuel and oxygen gases. 
It is very important to 
have such a homogeneous 
mixture of oxygen and fuel ; 

'mixed ! 
fuel and * 

fine flame 
! "33 drill for flame 

Fig. 13. 

7 These torches are obtainable from the Linde Air Products Company, 
30 East 42nd Street, New York City. 


otherwise the torches would blow themselves out. The type 
of orifice used is illustrated in Fig. 13. Fig. 14 shows the 

distribution of temperature 
in the oxyacetylene flame 
and also in the carbon arc. 8 

A furnace is required to 
heat objects to higher tem- 
peratures than those that are 
obtainable with torches. Gas 
furnaces for use in the labo- 
ratory are shown in Figs. 15 
and 16. 

Oxygen-gas furnaces can 
be made to yield very high 
temperatures; for example, 
Podszus and von Warten- 
burg, Linde, and Jung have 
described furnaces with a zir- 
conium dioxide tube using 
illuminating gas or oil vapor 
as fuel. 9 These furnaces at- 
tain temperatures of about 
2600°C. 10 

Electric furnaces. Electric furnaces for temperatures to 
500°C, useful for such applications as the baking out of 
charcoal traps, can be made by winding a coil of Nichrome 
or Chromel wire on an iron tube as shown in Fig. 17. The 

8 Flame and carbon-arc temperature: 

Kautny, Th., Leitfaden fur Azetylenschweisser, page 86. Halle: Marhold, 

Mathiesen, W., Untersuchungen uber den elektrischen Lichtbogen. Leipzig: 
Haberlandt, 1921. 

9 Podszus, E., Zeit. filr angew. Chem., SO, 17 (1917), 32, 146 (1919). 

von Wartenburg, H., Linde, H., and Jung, R., Zeit. fur anorg. u. allgem. 
Chemie, 176, 349 (1928). 

10 For treatment of high-temperature refractories see Swanger, W. H., and 
Caldwell, F. R., Bureau of Standards J. of Research, 6, 1131 (1931). 

Many of the high-temperature refractories are obtainable from the Foote 
Mineral Company, Philadelphia, Pennsylvania. 

See Langmuir, I., "Flames of Atomic Hydrogen," Indust. and Engin. Chem,., 
19, 667 (1927). 


carbon arc 

Fig. 14. 



tube is first covered with a piece of mica or asbestos sheet to 
avoid shorting out the winding. A simple way of fastening 
the ends of the winding is illustrated in Fig. 18. Various 

Meker burner 
enters furnace 


Fig. 15. 

firebricKs *s^ 

types of insulation may be used. For example, the inner 
tube and its resistance wire winding may be covered with 
several layers of asbestos. The furnace is assembled with 
transite 11 ends, using Insa-lute cement. It is necessary to 
avoid contact between the 
Insa-lute cement and the fur- 
nace wire at elevated temper- 

Nickel wire is suitable for a 
furnace winding. However, 
its resistance changes ap- 
proximately twofold when it 
is heated from room tempera- 
ture to 500°C. This behavior 
is in contrast to the behavior 
of the nickel-chromium alloys, 

Fig. 16. An improvised furnace. 

11 Transite is an asbestos fiber and Portland cement mixture formed under 
high pressure- into dense, monolithic sheets of high strength, rigidity, and dur- 
ability. It may be purchased from the Johns-Manville Corporation, 22 East 
40th Street, New York City. 




or ■ 



lo turns 

per inch 

tr an site 

whose change of resistance is negligible. The change of 
resistance of nickel may or may not be desirable; it may be 

desirable to have a large co- 
efficient if the resistance is to 
be used for regulating the 
temperature of the furnace. 

Electric furnaces which 
operate in air to 1100°C. 
may be made with the nickel- 
chrome alloys as resistors, a 
porcelain, Alundum, quartz, 
or magnesia tube being used 
to support the winding. Di- 
atomaceous earth makes an 
excellent insulator. 12 A use- 
ful furnace construction for 
the laboratory is illustrated 
in Fig. 19. 

Platinum may be used as resistor for temperatures greater 
than 1100°C., when it is desired to have the furnace operate 
in air. This resistor will operate up to a temperature limit of 
1600°C. In order to obtain a furnace temperature as near 
this limit as possible, Orton and Krehbiel used a Chromel 
"booster" winding on a tube 
mounted outside of and con- 
centric with the platinum 
winding. 13 The platinum wire 
may be wound on quartz glass, 
which has a temperature limit 
in air of 1300°C, on unglazed 
porcelain, for which the limit 
is 1400°C, or on clay, with 
a limit of 1700°C. However, 

about 50 volts heats this 
furnace to 500'C. 

Fig. 17. 

method of securing ends 
of winding 

Fig. 18. 

12 This may be obtained from Johns-Manville under the trade name Sil-O 
Cel. The calcined diatomaceous silica comes as a coarse granular material 
and as molded insulating bricks. 

13 Orton, E., Jr., and Krehbiel, J. F., Amer. Ceramic Soc, J., 10, 375 (1927). 


best of all is an Alundum tube (alumina with clay binder). 
Its limit, 1900°C, is above that of the platinum. 
Silicon is formed from 

Sil-O-Cel bricjs 




quartz or porcelain in a 
reducing atmosphere, and 
silicon attacks platinum. 
Accordingly, it is best to use 
a platinum-wound furnace in 
an oxidizing atmosphere. 
If, however, the wire is 
wound on an Alundum tube, 
it may be operated in a re- 
ducing atmosphere. 

Molybdenum or tungsten can be used as a resistor in an 
atmosphere of hydrogen; the limiting temperatures attain- 
able are 2200°C. and 3000°C. respectively. As a support 
for the resistor winding, Alundum can be used to 1900°C, 
magnesia to 2200°C, zirconia to 2500°C, and thoria to 

Fig.' 19. 


hydrogen flames 

to power supply 



wire -60-mil 


spaced '/4 

apart «* 




base plug 4 

welded or riveted 
sheet- iron case 

welded to 
base plate 

centering ring welded 
to base plate 

Fig. 20. Hydrogen furnace. 



3000°C. Porcelain is unsuitable, for the reason given above; 
namely, hydrogen blackens it at high temperatures. 14 A 
tungsten (or molybdenum) furnace is shown in Fig. 20. 
Most refractories cannot be subjected to high temperatures 

to vacuum 

water cooling, 
(topper tubing) 

water level' 
fiber seals 

filled with 
S»IO Cel) 




crucible on 
an Alundum 


mica window ^ 

with lead gaskets 

leads to 
power source 

over f Jow 
lead gaskets 
I copper clamp 

clamps to 

copper clamp 
(similar to 
that above) 

copper socket 

Fig. 21. The Arsem furnace. 

in vacuum because they either evaporate or are reduced by 
the vacuum (oxygen formed by dissociation is pumped 

Carbon and graphite tube furnaces can be operated to a 

14 For tables of physical and chemical properties of refractories see Hougen, 
O. A., Chem. and Met. Eng. t 30, 737 (1924). 


Above this temperature 

Sil-O-Cel brick cover 


Fig. 22. 

temperature of 2000°C. in vacuum, 
the carbon begins to vaporize, 
and at 2500° C. the rate of 
evaporation is rapid. In hy- 
drogen or nitrogen the tem- 
perature limit is 2000°C. At 
this limit chemical action be- 
tween the carbon and the gas 
sets in. However, in an atmos- 
phere of carbon monoxide, 
carbon resistors may be used 
at temperatures over 3000°C. 
Af urnace designed byArsem, 15 
which may be operated either 
in an atmosphere of carbon 
monoxide or in vacuum, is 
shown in Fig. 21 . This furnace 
has its resistor tube cut into a spiral to increase its resistance 
and flexibility. Connections are made to the ends of the 
horseshoe magnet to "blow" resistor tube with water-cooled 

copper jaws. 

Carbon grain resistors such 
as the one shown in Fig. 22 
have a higher electrical 
resistance than solid carbon 
and are often useful in the 

A carbon-arc 
shown in Fig. 23. 

Fig. 24 shows 
used for melting 
vacuum by heating with 
high-frequency current. 
It is peculiar to this 
method that the metal charge 
is at a higher temperature 

_ arc into crucible 

furnace is 

metals in 

stand for 


Fig. 23. After W. Schuen. 

]5 Arsem, W. C, Am. Electrochem. Soc, Trans., 9, 153 (1906). 


water cooling 

water cooling" 

to source of 
cycle A.C 

=3+ JS^. 

to vacuum l\piccm seals 

nurnn ^"^^. - 



copper tube silver^ 
soldered aroun d rim 
detail of cap 

Alundum spacer 
vitrosil tube 

detail of vacuum cell 

Fig. 24. Apparatus for melting metals in vacuum with high-frequency 


Fig. 25. Air bath. A Liebig condenser with a narrow cooling chamber is 

most efficient. 


than the crucible, a fact of practical value when working 
with extremely refractory metals. 16 

Fixed temperatures. Constant temperature may be 
maintained at 0°C. with melting ice, and at the boiling 
temperature of water by means of a device such as the one 
illustrated in Fig. 25. Other liquids and solids may be used 
for maintaining other constant temperatures; for example, 
a temperature of 444.6°C. is obtained by boiling sulphur. 
Some of the fixed temperatures useful for the calibration of 
thermometers and thermocouples are given in Table VI. 


Standard Temperatures 



Carbon dioxide (sublimation temperature) 

Mercury (melting temperature) 

Ice (melting temperature) 

Na 2 S0 4 • 10H 2 O (transition temperature) . 
MnCl2 • 2H 2 (transition temperature) . . . 
Steam condensation (at 760 mm pressure) . 

Naphthalene (boiling temperature) 

Tin (melting temperature) 

Benzophenone (boiling temperature) 

Sulphur (boiling temperature) 



Thermostatic devices. Here we cannot treat all of the 
many devices described in the literature for controlling the 
temperatures of furnaces and thermostatic baths. 17 How- 

16 Northrup, E. F., Frank Inst, J., 195, 665 (1923). 

Equipment for high-frequency heating is obtainable from the Ajax Electro- 
thermic Corporation, Trenton, New Jersey. 

17 Haagn, E., E. T. Z., Ifi, 670 (1919). 

Haughton, J. L., and Hanson, D., Engineering, 104, 412 (1917). 
Haughton, J. L., Journ. Sci. Instruments, 9, 310 (1932). 
Roberts, H. S., /. O. S. A. 6, 965 (1922). 
White, W. P., and Adams, L. H., Phys. Rev., U, 44 (1919). 
The Fish-Schurman Corporation is United States agent for German thermo- 
stats covering the temperature range — 35°C. to 300°C. 
See also the following: 

Beattie, J. A., Rev. Sci. Instruments, 2, 458 (1931). 
Roebuck, J. R., Rev. Sci. Instruments, 8, 93 (1932). 

Concerning the use of the thyratron for temperature control see the following: 
Hull, A. W., Gen. El. Rev., 32, 213, 390 (1931). 

Zabel, R. M., and Hancox, R. R., Rev. Sci. Instruments, 5, 28 (1934). 
Zabel and Hancox were able to get a constant temperature of 880°C. ± .06°. 


ever, the principle on which they operate is the same, namely, 
the balancing of the heat input to the furnace against its 
heat losses. The heat input is controlled by a pilot indicator 
which is continually kept oscillating about a mean position 
corresponding to the desired temperature. When it is on 
either one or the other side of the mean position, it modulates 
the heat input: When the pilot indicates the temperature 
low, the heat input is automatically increased, and when it 
indicates high, the heat input is diminished. In this sense 
one does not maintain a constant temperature but a periodic 
one which varies between more or less fixed limits about a 
mean temperature. 

As an example of a temperature-regulating device, let us 
consider a furnace with its winding made one arm of a self- 
balancing Wheatstone bridge, the bridge current in this arm 
serving at the same time as furnace heating current. The 
furnace winding must be made of nickel, molybdenum, 
tungsten, or platinum for this type of regulator, since the 
nickel-chromium alloys do not have a suitable temperature 
coefficient of resistance. The other resistances in the 
Wheatstone bridge may be rheostats made from a low- 
temperature-coemcient alloy such as constantan. The 
bridge galvanometer serves as the pilot to control the heating 

Let us compare this method with one which employs a 
thermocouple inside the furnace as a pilot. We see that 
there will be more lag between the time the heat input is 
altered and the time it affects the thermopile. As a result, 
with the thermocouple pilot the limits of the fluctuation of 
the furnace temperature are separated more than they are 
when the resistance of the heater wire serves as the pilot. 

Even when the furnace heater wire serves as the pilot, 
there are fluctuations due to the period of auxiliary instru- 
ments. These temperature fluctuations may be diminished 
simply by interposing alternate shells of thermal "ballast" 
and insulator between the furnace winding and the region 
that is to be kept at a constant temperature. The tempera- 


ture diffusion through such alternate shells is slow. The 
furnace tube itself, which separates the heater wires from 
the constant temperature region within, is usually adequate 
for this, because of its relatively low diffusivity, h; for 
example, one may obtain temperatures constant to about 

Fig. 26. 

0.01°C. inside the furnace tube even when the period of 
temperature oscillation of the furnace wiring outside is of the 
order of 30 seconds. 

The device shown in Fig. 26 is convenient for temperature 
regulation. 18 The two bulbs of this device have equal 
volumes, and they are equipped with identical nickel- 

18 Proctor, R. F., and Douglas, R. W., Journ. Sci. Instruments, 9, 192 (1932). 



heater to be kept at a 
constant temperature or 
"feeler" resistor whose 
resistance varies with the 
ambient temperature 

chromium alloy heaters and connected electrically as shown 
in the diagram, Fig. 27. The bulbs are filled with air and 
the pressures on either side of the mercury column are such 
as to hold the top surface of the mercury at the level of the 

tungsten contact when the 
voltage drop over the left 
resistance (see Fig. 26) is 
the same as the voltage 
drop over the right re- 
sistance. These voltage 
drops are equal when the 
temperature-sensitive feeler 
resistance is the same as 
the fixed constantan re- 
sistance. (See Fig. 27.) 
These resistances are ad- 
justed to be equal at the 
desired temperature. If the 
temperature of the feeler re- 
sistance is too high or too 
low, the heating in the two 
bulbs is unequal, and the 
resulting change in pressure 
in the bulbs opens or closes 
the mercury contact, and 
this in turn operates a relay 
actuating the heating and 
bridge current. 19 

The resistance used to op- 
erate the regulating device 

Fig. 27. 

may be either the heater resistance or it may be separate from 
the heater. In the latter case this arrangement is suitable for 
maintaining a constant temperature in a room. The feeler 
resistance is strung back and forth near the ceiling of the 

19 Mercury thermoregulators, relays, and electric bath heaters are handled 
by American Instrument Comnanv. 774 Girard Street, N. W., Washington 
D. C. 


room (at about 8 feet above the floor). For such an appli- 
cation the heaters, which the regulating device controls, are 
situated in front of the ventilator 
air inlet to the room. 

Thermostat baths use water for 
ordinary temperatures, oil or eutec- 
tic salt mixtures for elevated tem- 
peratures, and alcohol for low tem- 
peratures. Beattie gives the compo- 
sition of two eutectic baths. (See 
Table VII.) These baths are useful 
in the temperature range above 
120°C. The lower limits of their tem- 
perature ranges overlap the upper 
temperature limits of mineral-oil 
baths (150° to 200°C.) and heavy 
cylinder oil baths (150° to300°C.). 

The temperature of a water bath 
is controlled by regulating the heat 
onput. A mercury-in-glass bulb 
with contacts coupled to a relay 
as shown in Fig. 18, Chapter X, is 
suitable for a bath heated elec- 
trically. The device shown in 
Fig. 28 is effective for controlling 
the temperature of a gas-heated Fig. 28. After w. Ostwald. 

High-Temperature Bath Fluids 



Mineral seal oil 

Heavy cylinder oil 

30% LiN0 3 by weight ] 

14% NaN0 3 by weight \ 

56% KN0 3 by weight J 

Beattie, Rev. Sci. Instruments, 2, 458 (1931) 

150 to 300 

120 to 500 


bath. The aperture through which the gas for the flame 
passes is regulated by the thermal expansion and contraction 
of the mercury. With these devices the fluctuations of the 
bath are about 0.1°C. 

Temperature measurement. Temperature is always 
measured practically by a measurement of some temperature- 
sensitive property, such as light emission, electrical resist- 
ance, length, volume, thermal e.m.f., and so forth. All 
physical properties which vary with temperature are possi- 
bilities for such a measurement, although some properties, 
like electron emission, are so strongly influenced by chemical 
impurities or by the past physical history of the thermo- 
metric substance that they are useless. 

Liquid-in-a-bulb thermometers depend upon change of 
volume with temperature for their readings. Among them, 
two are of unusual interest. One, which was manufactured 
in Germany at one time, used gallium as liquid and fused 
quartz for the bulb and capillary. This thermometer was 
useful up to a temperature of about 1000°C, in contrast to 
the mercury thermometer, which is ordinarily useful only to 
200°C. However, with a high pressure of nitrogen (up to 
40 atmospheres) mercury-in-glass thermometers may be 
heated considerably above 200°C. A graphite thermometer 
with molten tin as the liquid has been made by Northrup. 20 
This thermometer may be used to 1680° without chemical 
reaction between the tin and the graphite. As tin does not 
boil at 1680°C, Northrup thinks that the limit in tempera- 
ture of this thermometer is probably several hundred degrees 
higher. The position of the tin in the graphite capillary is 
determined by a tungsten feeler. For gallium the tempera- 
ture range from the melting point to the boiling point is from 
29.7° to 1600°C., and for tin it is from 231.8° to 2260°C. 
The operation of the two thermometers described above 
depends upon these unusually long temperature intervals 
between the melting points and boiling points. 

20 Northrup, E. F., Pyrometry, page 464. New York: published by the Am. 
Inst, of Mining and Metallurgical Engineers at the office of the secretary, 1920. 



Thermocouples operate by virtue of the temperature 
dependence of the thermal e.m.f. generated by two sub- 
stances in contact. The thermocouple may be employed in 
the laboratory for temperature measurement from liquid air 
temperatures to the melting temperature of molybdenum. 

tap here to sift aluminum 
oxide down 

screen -Z5 

rubber stopper 

pure powdered AI2O3 
(free from potassium) 
for white sapphires — 
"scientific brilliants" 

12% titanium oxide 
>lue sapphires 
chromic oxide 

water . 


< fused Al 2 0i base 
platinum tube carrier 

iron rod support 
iron shield 

rack and pinion for 
vertical adjustmej-it 

horizontal adjustment 
by sliding base 
on table > — " 

Fig. 29. Veraeuil's arrangement for making artificial rubies and sapphires. 
Verneuil, A., Ann. de Chemie et Physique, S, 20 (1904). 



The base metals commercially available are commonly 
used as thermoelectric wires. Chromel-Alumel wires have a 
high coefficient of thermal e.m.f. They are obtainable from 
the factory matched to give the temperature to ± 5°C. 
Copper and constantan wires also have a high thermal e.m.f. 
These wires have the advantage over Chromel-Alumel that 
they are easily soldered. The LeChatelier combination 
(platinum and 10 per cent platinum-rhodium) is used for 
precision measurements. Special thermocouple metals, such 
as tungsten, molybdenum, and their alloys, are useful at very 
high temperatures. 

It may be desired to calibrate a particular thermocouple 
with fixed standard-temperature baths such as those listed 
above in Table VI. The best procedure is to use the cali- 
bration curve supplied by the factory, which gives the e.m.f. 
at frequent temperature intervals, and to plot an empirical 
correction curve for it from the calibration data. 

Radiation pyrometers determine temperature by the 
measurement of light emission. There are several types, and 
descriptions of them and their operating characteristics 
appear in many books. The type in most common use 
measures, with a special photometer, the intensity of mono- 
chromatic light (6600 A) emitted by the incandescent body 
whose temperature is being measured. 

Table VIII is useful for estimating the temperature of a 
body from its color. 


Color Temperatures 


Approximate Temperature 


Incipient red heat 

500 to 550 

Dark red heat 

Bright red heat 


Yellowish-red heat 


Incipient white heat 


White heat 



Notes on the Materials of Research 

Alkali metals. One of the alkali metals may be required 
for the sensitive surface of a photocell or as a thin-film filter 
for ultraviolet light; or in the vapor phase the metal may be 
used for the demonstration of the phenomenon of resonance 
radiation. For these and other applications we will outline, 
briefly, some of the ways of manipulating these very reactive 

The alkali metal may be prepared from the alkali chloride, 
reduced with calcium metal in an evacuated glass tube : 

2MC1 + Ca -> 2M + CaCl 2 . (1) 

The reaction progresses in the indicated direction at elevated 
temperatures on account of the removal of the free alkali 
metal, M, by evaporation. This reaction may be varied : A 
chromate of the alkali metal may be used instead of the 
chloride, and zirconium metal may be used instead of cal- 
cium. The reaction applies to the preparation of all the 
alkali metals, with the exception of lithium, which reacts 
with the glass or quartz; lithium is best reduced from its 
chromate with zirconium metal in an iron apparatus. 

We will consider, in detail, how potassium may be pre- 
pared by the reaction indicated in Eq. 1. Pulverized 
potassium chloride and calcium metal filings are mixed to- 
gather in a closed-end iron tube in stoichiometrical propor- 
tions (3.7 g KC1 to 1 g Ca). This iron tube is introduced 
into the thickened end of a hard-glass tube as shown in 
Fig. 1. The glass is thickened in order to allow the attain- 
ment of the maximum temperature; at lower temperatures 
where a thinner glass wall would collapse the reaction pro- 




ceeds very slowly. After the iron tube is introduced, the 
hard-glass tube is closed by fusing the glass with the hand 
torch. After a good vacuum is attained, the chemicals are 
heated, slowly at first and finally strongly until the reaction 
is complete. The chemicals may be heated until the glass 
starts to soften, but too much heat should be avoided, since 
it will distill calcium metal. The reduced and once-distilled 
metal condenses in the bend of the tube as shown in Fig. 1. 
From there, it is worked down with the flame into the 
receiving ampoule, where it is sealed off as illustrated. 

Metallic potassium first condenses here 
This point is later heated to 
cause the potassium to flow 
down into the ampoule. / / 

Iron tube 

calcium fmngs 

and powdered 




Fig. 1. 

The alkali metals react vigorously with air, and the 
ampoule should be opened without exposing the metal to 
air. This is done by the following procedures : The ampoule 
is constructed of Pyrex glass with an annular tungsten ring 
to spring the glass. After the ampoule is mounted in the 
vacuum system, the tungsten ring is heated with a high- 
frequency induction coil until the glass breaks. (See Fig. 2.) 

A scheme which does not involve the use of high-frequency 
heating but which breaks the glass by impact is illustrated by 
Fig. 3. The illustrated depression in the tube wall acts as 
a safety to confine the armature and prevent accidental 



fracture of the ampoule until the apparatus is sealed onto 
the high-vacuum system. During this sealing operation the 
depression is blown out of the 

way of the armature. The 
armature is operated in the 
vacuum, by means of an ex- 
ternal electromagnet, to break 
the tip of the ampoule, thus 
exposing the alkali metal. 
The tip may be scratched 
with a file to facilitate break- 

metal fingr 
to break 
ampoule C 
when ^ 

heated by^ 
high-fre- *^ 

induction J? 


=^ ^* high-fre- 

^ * L <-=*=* =7 induction 


filled with 
an alkali 

to vacuum 
Ay item 

The ampoule may be cooled 
in a beaker with dry ice in 
the bottom and carbon di- 
oxide vapor above. The tip is opened under the surface of 
the carbon dioxide vapor. The ampoule is then quickly 
transmitted to the vacuum system, sealed in, and evacuated. 
The expansion of carbon dioxide in the ampoule through 
the tip prevents access of air to the alkali metal. 

safety notch to prevent pre mat ore 
breaking of ampoule ttp 

This notch is blown out when 
the apparatus is sealed to 
the vacuum system 

fQj3EB5SfBE3$ )j: 

iron armature sealed into 
a glass tube to prevent its 
out gassing This armature, 
when actuated by a hand 
magnet, breaks the tip of 
the ampoule. 

Fig. 3 

seal hereto 
system >i 


alkali metal 

The alkali metals, as obtained commercially in small cubes 
or irregularly shaped pieces, are packed submerged under 
kerosene. The metal may be cleaned and manipulated as 



follows i 1 First the metal is washed in dried petroleum ether 
or benzene to free it from kerosene. The petroleum ether or 
benzene is dried by shaking it in contact with calcium chlo- 
ride. (Carbon tetrachloride or chloroform should not be used 
to wash the metal, since an explosive compound is formed.) 
The metal is then fused in the bottom of an 8-mm glass tube 
and sucked up into a 1-mm capillary glass tube with a rubber 
hose. This 1-mm tube is sealed off with a flame just above 
the metal. At the other end the alkali is protected from 
the air by soft wax. A suitable length of this composite 

scaled to vacuum 


The metal is distilled from the ampoule A 
to C , sealed off at B ; distilled from C 
to E, sealed off at "D ; etc., until the 
purified metal is finally distilled from I 

alkali metal into the vacuum system, 

in capillary tube 

Fig. 4. 

glass-metal rod may be cut off with wire cutters and intro- 
duced into a distilling bulb fastened to the vacuum system 
where the metal is desired.. (See Fig. 4.) 

A distillation procedure 2 for sodium metal is illustrated in 
Fig. 5 whereby the metal is renuxed under vacuum to free 
it of hydrogen and carbohydrates (the hydrogen contained 
in potassium or sodium, measured as a gas at atmospheric 
pressure, may amount to one or two hundred times the 
volume of the metal). The metal cubes are washed to free 
them of kerosene, as described above, and then they are in- 
troduced into Chamber I. After the whole system is evac- 
uated, the metal is fused in this chamber. Chamber I acts 
as a separating funnel. The fusion is accomplished by the 

1 Wood, R. W., Phys. Rev., U, 353 (1933). 

2 1 am indebted to Dr. Carl F. J. Overhage for this procedure. 



application of a soft flame, so that the metal runs into 
Chamber II, leaving the dross behind. Chamber I is then 
removed at the seal-off. The metal is heated in Chamber II 
with a small electric furnace. Here it is refluxed for several 

to vacuum system 

heater similar 
to that below 

alkali metal 

first seal-off 

about 8 turns per 1 
inch of *24 Chromel 


10-mm glass tube 

Fig. 5. 

hours. The distilled metal condenses in the asbestos-insu- 
lated tube above Chamber II. This refluxing allows hydro- 
gen and hydrocarbon vapors to be pumped away. After 
this treatment th^ metal is distilled into Chamber III by a 
heater wire around the condenser tube. Chamber II is then 
removed at the seal-off. Chamber III may be the receiver 
for the metal, or it may be further refluxed and distilled into 
a final receiving ampoule. Electric heat is recommended for 



distilling the alkali metal, since there is some danger of 
breaking the glass if it is heated with a torch. 

In manipulating the alkali metals the following precau- 
tions should be observed : The amount of metal manipulated 
should never be greater than necessary. A box of sand 
should be at hand for the control of accidental fires. The 
alkali metals should never be allowed to come in contact 
with water. Used metal and apparatus containing the alkali 

air blast to 
cool top of bulb 

iron Kettle, 
heat is applied to 
side of hettle to 
avoid forming" a ge^se* 

Fig. 6. 

metals should be disposed of by burying only. It is advis- 
able to wear goggles to protect the eyes while manipulating 
the alkali metals. 

Sodium may be prepared by electrolysis through the soda- 
glass walls of an electric lamp. A 32-volt lamp, which has a 
larger tungsten filament wire than the 110- volt lamp, is best 
for this purpose. The lamp bulb is first evacuated by means 
of a side tube sealed on for this purpose. It is then dipped 
in a bath of fused sodium nitrate and nitrite and connected 
to a source of electrical energy as shown in Fig. 6. Current is 
carried from the tungsten filament to the glass walls of the 
lamp bulb by electrons or by means of a sodium discharge, 
or in special cases by means of an argon discharge. The 



practical details of this procedure are due to Dr. R. C. Burt s 
who graphically described the procedure as one which allows 
the vacuum to be electroplated with sodium. 3 The free 
metal is formed from the reduced sodium ions (which migrate 
through the solid glass electrolyte when a current flows). 
These ions are reduced by electrons, or negative sodium 
(or argon) ions. Faraday's law applies to the electrolysis. 
The spectrum of the sodium vapor discharge has been photo- 
graphed and the spectrum indicates high purity of the 
electrolyzed metal. Impurities were estimated by Dr. Burt 
as being present, at most, in the proportion of 2 parts per 



' 1 dSh9 B?^s 

eiiitfmaEc- •■".- laltn 


Fig. 7. Fluoresence of sodium vapor. 

million. Sodium prepared by electrolysis is characterized 
by the fact that it is completely free of hydrogen and carbo- 

The electrolysis current varies from a few milliamperes 
when the current is carried entirely by electrons to a few 
hundred milliamperes when it is carried by sodium ions. The 
sodium discharge is obtained by simply removing the air 
blast on the lamp bulb which normally serves to keep the 
metal condensed. 

Burt states that the spectrum from the sodium discharge 
is not reversed; a warmed lamp containing sodium will 

3 Burt, R. C., /. S. A., 11, 87 (1925). 



fluoresce if the light of the sodium discharge from another 
lamp is focused on it. (See Fig. 7.) 

Sodium may be introduced into quartz photocells by 
means of a graded seal as shown in Fig. 8. 

Potassium can be electrolyzed through a potassium glass 
which is free of sodium and lead. A bath of fused potassium 
nitrite and nitrate is used. 

The alkali metals potassium and sodium may be dissolved 
in the volatile solvent, liquid ammonia, and deposited where 

tungsten electrodes 

scale of 

area o"f photo- 
cell bulb to be 
coated with sodium 

coating of 
sodium to be 
transferred to 
the photocell 

Fig. 8. 

they are desired by boiling away this solvent. Lithium is 
managed in a similar manner with aethylamine as solvent. 

All the alkali metals react with glass at elevated tempera- 
tures and especially with lead glass, with which they should 
not be allowed by come in contact. 

The resistance of Pyrex-glass tubes toward sodium can be 
improved if they are lined with a film of borax or boracic 
acid. The tube to be lined is filled with a hot saturated 
solution of borax. The borax precipitates from this solution 



as crystals on the inner glass walls of the tube as the solution 
cools. When the glass has become lined with a thin coating 
of crystals, the solution is drawn off and the tube carefully 
dried. It is then evacuated, and the water is driven off by 
heating. At first the heating is gentle, but finally the tube 
is fired at the maximum temperature the glass will stand. 
This gives the tube a smooth sodium resistant inner surface. 

The potassium-sodium alloys, lying within the composi- 
tion range 45 to 90 per cent potassium, are liquid at room 

Alkali-earth metals. The chief uses of the alkali-earth 
metals, as getters, depend on their reactions with oxygen to 
form oxides, with carbon dioxide to form carbides and 
oxides, with water to form hydrides and oxides, and with 
nitrogen to form nitrides. 

When fresh calcium filings are heated in a quartz tube or 
heavy-walled Pyrex side tube, connected to an apparatus 
such as a thermopile, the calcium reacts with all the re- 
sidual gases (except the noble gases). A fairly good vacuum 
can be obtained with such a side tube even when start- 
ing at atmospheric pressure. For example, the argon 
spectrum may be obtained in a discharge tube evacuated 
from half of an atmosphere pressure with such a calcium 
side tube. Each time the tube is evacuated from atmos- 
pheric pressure with calcium, the residual pressure of 
argon (calculated from its abundance in the atmosphere) 
is increased by 7 mm. 

i Barium is a more reactive metal than calcium. 4 It is used 
as a getter for commercial radio tubes. For this applica- 
tion, the metal is sometimes cast in a seamless tube of nickel 
or copper which is drawn down to wire. These composite 
wires are known as Niba and Cuba wires. The wires are cut 
into short lengths, which are introduced into radio tubes and 
other places where the getter action is desired. The volatile 

4 Barium and strontium metal of a guaranteed purity of 99.5 per cent may- 
be purchased from the Varlacoid Chemical Company, 15 Moore Street, New- 
York City. 



core metal is subsequently boiled out of the nickel or copper 
covering tube by means of heat generated with a high- 
frequency induction coil. 

Mercury. Although mercury approaches the noble metals 
in chemical inertness, it is easily contaminated, especially by 
other metals. This is because, as a liquid, it is a fairly good 
solvent. A simple test for the purity of a sample of mercury 

The scum 
is removed 
by strain-, 
ind through 


in filter 



are re- 
by pass 
ing the 



&Q% H3 S0 4 . 

By bubbling air 
through the mercury 
many of the dissolved 
metals arc converted 
into insoluble oxides 
which form a scum 
on the surface. 

The noble metals 
and "tin arc re- 
moved by vacuum 

is to raise a clean glass rod slowly up through the metal 
surface. If the mercury is clean, the glass will come up 
without any adhering mercury droplets. 

The contaminations commonly found in mercury may be 
classified according to the manner in which they can be easily 
removed. First come surface contaminations by materials 
which do not dissolve in the liquid metal and may, accord- 
ingly, be removed by filtering the metal through pinholes in 
filter paper or through a chamois skin. Second, there are the 
dissolved metals. Those which are oxidizable are first con- 


verted to insoluble oxides by the blowing of air through the 
mercury as shown in Fig. 9(a). The oxides form a scum on 
the mercury surface and may later be filtered off. Mercury 
is practically free of impurities of this type if, after air has 
been blown through the liquid metal \ hour, no scum has 
formed on the surface. The alkali metals fall into this class 
of impurities; here also belong zinc, with a high vapor 
pressure, and copper and lead, with low vapor pressures. 
These metals, which are more reactive than mercury, can also 
be removed by exposing the mercury to a solution of 10 per 
cent HN0 3 or 80 per cent H 2 S0 4 . This is shown in Fig. 9 (b) . 
Thirdly, there are the dissolved metals, such as the noble 
metals and tin, which cannot be removed by oxidation or 
acid. Copper and lead may also be considered as belonging 
to this class of contaminations. These metals are removed 
by vacuum distillation of the mercury at a temperature of 
about 180° to 200°C. (at which temperature the mercury 
distills at the rate of approximately \ g/cm 2 /sec.) as indi- 
cated by Fig. 9(c). 

The vapor pressure of mercury is given in Table I. It 
is to be noted thoughtfully that at room temperature the 

Vapor Pressure of Mercury 


Vapor Pressure 










vapor density of mercury is many times greater than the 
accepted nonpoisonous concentration limit, which is 1 milli- 
gram of mercury per cubic meter. According to Stock, 
continual breathing of air containing only 15 micrograms 
per cubic meter of mercury for a few weeks will make most 


persons ill. 5 The vapor pressure of mercury is hazardously 
high in many laboratories. In a Berlin physical laboratory 
the typical concentration of mercury vapor in the air was 
found to be about 20 to 60 micrograms per cubic meter; in 
one room it was 500 micrograms per cubic meter. Heat 
produced by turning on mercury pumps doubled the concen- 
tration of mercury vapor in the air. 6 

Platinum metals. Platinum is chemically resistant to 
alkalies and hydrofluoric acid. However, it is attacked by 
chlorine vapor and aqua regia. Metallic salts should not be 
heated in platinum under conditions which may result in the 
reduction of the metal and the consequent debasement and 
embrittlement of the platinum. This applies particularly to 
lead salts. The elements phosphorus and silicon also attack 
platinum and make it brittle, and they may change its other 
properties. For example, even the small amount of silicon 
introduced into the platinum when it is heated in contact 
with porcelain in a reducing atmosphere makes an appre- 
ciable change in the thermoelectric power and electrical 

Platinum is so ductile that wires may be drawn directly as 
fine as 20/z diameter. By Wollaston's procedure a platinum 
rod is covered with a close-fitting silver tube, and this com- 
posite rod is drawn through wire dies. After the silver has 
been etched off the final wire with nitric acid, the platinum 
wire obtained may be as small as \\x in diameter. Wollaston 
wire is often used for fuses to protect delicate instruments. 7 

A physical property of platinum which is of interest to 
the physicist is its " transparency " to hydrogen gas at tem- 
peratures above 700°C. (See Fig. 10.) This property is 
employed to obtain very pure hydrogen. 

5 Stock, A., and Cucuel, F., Ber. deutsch. diem. Ges., 67, 122 (1934). 

6 Miiller, K., and Pringsheim, P., Naturwiss., 18, 364 (1930). See also 
Turner, J. A., Pub. Health Bull., 39, No. 8 (1924). Goodman, Clark, "Mercury 
Poisoning, A Review of Present Knowledge," Rev. Sci. Instruments, 9, 233 

7 Wollaston and Taylor process wires are handled by Baker and Company, 
54 Austen Street, Newark, New Jersey. 

Littelfuses are obtainable from radio supply houses. 



Platinum is a refractory metal. For this reason it may be 
used for furnace windings and as a base for oxide cathodes. 

Iridium is harder and more resistant to chemical attack 
than platinum; it is not attacked by aqua regia. Accord- 
ingly, it is often alloyed with platinum in proportions up to 
30 per cent to yield a metal which is superior to platinum in 
respect to chemical resistance and hardness. 

Rhodium is alloyed with platinum (90 Pt to 10 Rh) to 
yield the LeChatelier thermocouple alloy. Rhodium is a 
bright inert metal and for this reason it is used for electro- 
plating other metals. 

Osmium is the most refractory metal of the platinum 
family, with a melting temperature of 2700°C. It was once 

200 *100 GOO 800 900 11 OO 1500 1500 170O 

temperature in degrees centigrade. 

Fig. 10. 

Borelius, G M and Lindblom, S., Ann. d. Physik, 82, 201 (1927). Smith- 
ells, C. J., and Ransley, C. E., Roy. Soc, Proc, 150, 172 (1935). Sieverts, 
A., Zeits.f. Metallkunde, 21, 37 (1929). 

used in incandescent lamps but has now been replaced by 
tungsten for this use. Incidentally, it is the heaviest known 
substance, having a density of 22.5 g/cm 3 . 

Palladium is the least noble of the platinum metals. It 
oxidizes when heated in air and is dissolved in nitric acid. 
Hydrogen diffuses through palladium more rapidly than 
through platinum. At atmospheric pressure palladium will 
dissolve about 6 mg H 2 per 100 g of metal to form the " alloy '" 
Pd 2 H. The hydrogen is given off again if the metal is heated, 



in vacuum, to temperatures above 300°C. (See Fig. 10.) 
This property affords a convenient source of extremely 
pure hydrogen in small quantities. 

The refractory metals : Tungsten, molybdenum, tantalum, 
and so forth. Tungsten is the most refractory metal and 
also the strongest. Wires of .0014 inch in diameter exhibit 
a tensile strength of 590,000 lbs. /square inch. Tungsten is 

quite "unorthodox" in its 
behavior with respect to cold 
working and heat. Passing 
it through dies makes it more 
ductile, while heating it to 
a temperature greater than 
1000°C. causes recrystalliza- 
tion and makes it brittle, a 
situation just opposite to the 
behavior of most metals. 
The ductility of tungsten at 
ordinary temperatures is due 
to its long fibrous crystal 
grains. Fig. 11 shows the 
ductility of tungsten at vari- 
ous temperatures. It will be 
noted that recrystallized brittle tungsten is ductile if heated 
to temperatures greater than 200°C. 

Traces of water vapor are corrosive oil the tungsten fila- 
ments in vacuum electric lamps. The water molecule reacts 
with hot tungsten to form tungsten oxide and atomic hydro- 
gen, both of which evaporate to the glass wall of the bulb, 
where, owing to catalytic effect of the glass, they react to give 
metallic tungsten and water vapor again. The water molecule 
is now free again to repeat its action on the tungsten filament. 
Tungsten reacts with oxygen and carbon monoxide, in 
vacuum, to form oxides and carbides. Tungsten is not 
attacked or affected by mercury vapor or hydrogen gas. In 
air, at a yellow heat, tungsten reacts with oxygen to form 
volatile oxides, which distill off as white smoke. 


temperature in degrees 

Fig. 11. 

Espe, W., and Knoll, M., Werkstoff- 
kunde der Hochvakuumtechnik, page 
18. Berlin: Julius Springer, 1936. 


Molybdenum is more ductile than tungsten. Otherwise, 
it is very similar to tungsten, and the two metals form alloys 
in all proportions. Some of these alloys are used com- 
mercially. Their properties are, in general, a compromise 
between the higher melting temperature of tungsten, on the 
one hand, and the greater workability and machinability of 
molybdenum, on the other. 

Molybdenum and tungsten do not soft-solder or amal- 
gamate with mercury, but both metals may be welded to 
nickel or Advance alloy. Nickel is frequently welded to 
tungsten to facilitate connecting it by spot-welding, solder- 
ing, or brazing to other less refractory metals. 

Tungsten or molybdenum may be cleaned by heating the 
metal to a red heat and rubbing its surface with a piece of 
potassium or sodium nitrite. 

In many respects tantalum is like molybdenum and 
tungsten. 8 Tantalum, when it is very pure, is one of the 
most ductile metals. However, when heated in hydrogen or 
air, tantalum becomes brittle. To anneal tantalum, it must 
be heated to about 800°C. in a vacuum better than 5 X 
10 -2 mm of mercury. Because tantalum readily gives off 
occluded gas if heated above 800°C, it is used as a construc- 
tion material in vacuum tubes. 

To spot-weld this metal successfully, it must be submerged 
under carbon tetrachloride or water. It may be machined 
using carbon tetrachloride as a cutting fluid, and spun using 
hard laundry soap as lubricant. 

Columbium occurs with tantalum and has many proper- 
ties in common with it. It is less refractory and more ductile 
than tantalum. It is used as a substitute for tantalum. 

Rhenium is the heaviest member of the manganese sub- 
group in the periodic table, and it is very refractory, its 
melting temperature being only about 200° below that of 

8 Tungsten, molybdenum, and tantalum may be obtained from the Fan- 
steel Products Company, Inc., North Chicago, Illinois, and Callite Products 
Company, 595 Forty-Ninth Street, Union City, New Jersey. 



Alloys. Invar. The iron-nickel alloy, 63.5 Fe, 36 Ni, 
0.5 Mn, is known as Invar. Its coefficient of expansion is 
only low for temperatures below 120°C, being 4 X 10~ 7 per 
degree centigrade. The heat conduction of Invar is also 
very low, being only ^o that of copper. Invar does not 
corrode. It is used for the construction of surveyor's tapes 
and instruments in which the dimensions are required to 
remain constant in spite of temperature changes. The alloy 
melts at 1425°C. 

Electrical-resistance alloys* Nickel-chromium alloys are 
characterized by a high electrical resistance (about 58 times 
that of copper), a low temperature coefficient of resistance, 
and a high resistance to oxidation. Examples are Chromel 

\Chromel X" 

ZOO 300 400 500 600 700 &0O 900 1000 1100 

Temperature — degrees centigrade 
These curves represent average values. Actual samples 
of the materials may depart from these values from 
negligible amounts at 20°C to as-much as ±4/a% 
at Hoof For precise work samples of the material 
to be used should be tested. 

Fig. 12. 

9 Chromel is manufactured by the Hoskins Manufacturing Company, 
Detroit, Michigan. Nichrome is manufactured by the Driver Harris Com- 
pany, Harrison, New Jersey. 


A and Nichrome V, of which the typical composition is 
80 Ni and 20 Cr, with the melting point at 1420°C. 

When some iron is added to the nickel-chromium alloys, 
it makes them more ductile. Nichrome and Chromel C are 
examples of these iron-containing alloys. The typical com- 
position of Nichrome is 60 Ni, 12 Cr, 26 Fe, 2 Mn, and of 
Chromel C, 64 Ni, 11 Cr, 25 Fe. The melting temperatures 
of these alloys are 1350° and 1390°C. respectively. The 
change of resistance with temperature for these alloys is 
illustrated in Fig. 12. 

Thermocouple alloys. Chromel P gives a useful base- 
metal thermocouple in combination with the alloy Alumel 
(94 Ni, 2| Mn, \ Fe). The thermocouple wires are welded 
together under a borax flux to make the junction. None of 
the Chromels braze, but they all may be welded to nickel. 

Constantan (45 Ni, 55 Cu) has practically zero tempera- 
ture coefficient of resistance up to a temperature of 400°C. 
Also, it gives a high thermal e.m.f . against copper, making an 
excellent thermocouple. Constantan exhibits high resist- 
ance to oxidation and corrosion. It solders easily. 

Solders. Solders are required to flow onto the surface of 
the metals to be joined and to alloy with the surface layers of 
the metals. Also, they should be ductile, have high strength, 
and be noncorrosive. 

Silver solder best meets all these requirements. It is used 
for joining brass, steel, stainless steels, and many other 
metals. Silver solders are, in effect, brazing alloys of the 
composition (4 Cu to 3 Zn) with silver added. A solder 
melting at 693 °C. contains 65 per cent silver, while one 
melting at 760°C. contains but 20 per cent silver. 

High-quality soft solder is half tin and half lead. Solders 
are often made with a higher content of lead, since the tin 
component is more expensive than lead. Such solders are 
inferior, since it is the tin component that makes the solder 
run well and adhere well. " Half-and-half " solder melts at 
188°C. The properties of various solders are given in 
Table II. 




Properties of Solders 






Soft solder: 

Wood's metal 

soft solder eutectic .... 

Bi 50, Cd 12.5, Pb 25, Sn 12.5 
Pb 36, Sn 64 
Pb 50, Sn 50 

Ag 45, Cu 30, Zn 25 

Cu 54, Zn 46 

Ag 20, Cu 3, Zn 2, Sn 75 




Hard solder: 

silver solder 

B C 

brazing compound. . . . 
Intermediate solder 



Composition of fluxes: 

A— (a) Flux: 40 ZnCl 2 , 20 NH 4 C1, 40 H 2 0. 

(b) Paste: 90 Petrolatum, 10 NH 4 C1. 

(c) Solution of rosin in alcohol. 

B — (a) Thin paste composed of water and 10 parts powdered borax and 
1 part boracic acid, 
(b) Borax applied dry. 
C — Handy flux. Manufactured by the Handy and Harman Company, 
Bridgeport, Connecticut. This is an excellent flux. It has a lower melting 
point than borax. 

Brass and bronze. Brass is the most widely used con- 
struction material in the physical laboratory. It is funda- 
mentally a copper-zinc alloy. Red brass (10 to 20 per cent 
zinc), or so-called Tombak alloy, is used for making flexible 
corrugated tubes (such as Silphon tubes) when maximum 
ductility is required; yellow or common brass, which con- 
tains copper and zinc in the proportions 65 to 35, with small 
lead additions to increase its machinability, is used where 
springiness is desired. 

Brasses are less expensive than the copper-tin alloys or 
bronzes. They are also softer and more ductile. Brasses 
are used for drawing and rolling, whereas bronzes are prima- 
rily casting materials. Bronze castings are much more likely 
to be vacuum tight than brass castings. Also, because 
bronzes have small crystals of the hard brittle compound 
CiuSn, they make good bearing metals (the 68.2 copper 


bronze, Cu 4 Sn, is the true speculum metal used for optical 
gratings and for mirrors). 10 

Dur aluminum. The aluminum alloy with composition 
95 Al, 4 Cu, ^ Mg, ^ Mn, is known as Duraluminum. Dura- 
luminum is employed extensively in many cases where brass 
was formerly used. For about 45 minutes after it has been 
heat treated at 530°C. and quenched in water, Duraluminum 
is ductile and can be rolled, bent, or cold-worked. After 
this interval a copper aluminum compound is precipitated 
out of solid solution, and this precipitate "keys" the crystals 
of the alloy at their slip planes, giving the alloy increased 
hardness and strength. The tensile strength, originally 
30,000 lbs. /square inch after quenching, becomes as great as 
75,000 lbs. /square inch after cold-working and aging. Dura- 
luminum rivets are frequently stored in buckets cooled with 
dry ice. They may be used as desired, for this low tempera- 
ture arrests the aging process, and the metal does not harden 
until after it has warmed up to room temperature. 

Wood. 11 Two kinds of wood are obtained from a tree: 
heartwood and sapwood. The heartwood is formed early in 
the life of the tree and, as the name implies, is found near the 
center of the trunk. Protoplasms present when the tree is 
young are gradually replaced by deposits of gum, minerals, 
tannin, and pigments to form this heartwood as the tree 
becomes older. These substances make it heavier, stronger, 
and in most cases darker than the sapwood. The heartwood 
of the redwood tree, which is particularly free from gums and 
oils, is an exception. In other heartwoods there are abun- 

10 Lord Ross' famous 60-inch mirror contains 70 Cu 30 Sn; an old Roman 
mirror contains 64 Cu, 19 Sn, 17 Pb; an Egyptian mirror contains 85 Cu, 
14 Sn, 1 Fe. Brady, G. S., Materials Handbook. New York: McGraw-Hill 
Book Company, 1931. 

u Fowle, F. E., Smithsonian Physical Tables. Washington: The Smith- 
sonian Institution, 1934. 

Koehler, Arthur, Properties and Uses of Woods. New York: McGraw-Hill 
Book Company, 1924. 

Marks, L. S., Mechanical Engineers' Handbook. New York: McGraw-Hill 
Book Company, 1930. 

"Mechanical Properties of Woods Grown in the United States," Department 
of Agriculture, Bull. 556. 


dant deposits. For example, in lignum vitae, these com- 
pounds produce an oiliness (especially when the wood is wet) 
which makes it suitable as a bearing material. 

Sapwood, or the outer part of the tree, is more pliable than 
heartwood. Therefore, in using such woods as hickory and 
ash, which are noted for their adequate strength, the outer 
part of the trunk may be preferred to the heartwood be- 
cause of its pliability. 

Effects of temperature. Some of the effects of temperature 
on wood are due to the gum deposits. High temperature 
softens these gums, making the wood weaker and more liable 
to split. On the other hand, low temperatures produce in- 
creased brittleness. 

The thermal expansion of wood in directions parallel and 
perpendicular to the grain is given in Table III. It will be 
noted that the expansion parallel to the grain is less for wood 
than for most of the metals. This property is a useful one, 
and it should be kept in mind and used in the construction of 
instruments where invariance of length is desired, as, for 
example, in a telescope tube, in which the relative distance 
between the optical components should not change with 
changes in temperature. 


Linear Expansion of Wood and Various Other Solids per Unit Length 
per Degree Centigrade X 10~ 6 





Cast brass 


Cast iron 




Glass, plate and crown 



Parallel Perpendicular 
6.3 48 


4.9 55 


5.4 34 


6.5 48 

Sugar maple 




The heat conductivity of several common types of wood 
is given in Table IV. The conduction of heat is from two to 
four times as great along the grain as it is across it. The 
conductivity depends, in a large measure, on the moisture 
content. To obtain maximum heat insulation, the wood 
must be dry. To keep it dry, particularly if the wood is to 
be exposed to low temperatures, it should be coated with 

Heat Conduction of Wood and Some Other Materials in C. G. S. Units 


Specific Gravity 

Heat Conductivity 





Hard maple 


White pine: 

along the fiber 


across the fiber 


Cotton — firmly packed 


Hair, felt 


Effects of moisture. One drawback to the use of wood as a 
material for construction, especially for scientific apparatus, 
lies in the fact that its dimensions may change considerably 
with its change in moisture content. We may take the 
shrinkage from the green to the dry condition as an index of 
the changes one may expect with changes in humidity and 
residual curing. This shrinkage (radial and tangential) for 
several woods is given in Table V. 


Shrinkage of Wood from the Green to the Dry Condition 





















4.2 to 14 

2 to 8.5 

Magnolia, evergreen 


Sugar maple 

Sugar pine 

Pine, northern white 

Range of all commercial woods 

Wood Handbook, United States Department of Agriculture, September, 1935 


Among the hardwoods, evergreen magnolia is prized as 
one which does not warp. This may be understood by re- 
ferring to Table V, where we see that of all the woods it is the 
one whose radial and tangential shrinkages are most nearly 

Most of the shrinkage in wood is at right angles to the 
grain; the longitudinal shrinkage, taken from the green to 
the cured condition, is seldom greater than T V to -J- per cent. 
(It is greater than this for some woods, particuarly woods 
grown under strong compression. Yellow pine compression 
wood, for example, may shrink longitudinally as much as 
2J per cent when it is cured. Redwoods also exhibit con- 
siderable longitudinal shrinkage. However, longitudinal 
shrinkage is negligible for most of the other woods.) This 
property of wood, in addition to the low thermal expansion 
parallel to the grain, explains why wood has been used so 
successfully for rulers ; it may suggest other applications for 
wood in the laboratory. 

When a piece of wood is carved or cut to precise dimen- 
sions that are to be maintained, it should be painted at once 
with several coats of shellac, in order to maintain the mois- 
ture equilibrium already established. Linseed oil is less 
effective, while paraffin is more effective than shellac for this 
purpose. Molten paraffin is applied by pouring it over the 
surface with a spoon. The boiling of wood in paraffin 
causes it to become brittle. 

Strength of wood. Strength and rigidity do not vary from 
wood to wood as much as is commonly supposed. For ex- 
ample, the bending strength of shagbark hickory is only 
2.6 times as great as that of sugar pine, and pine is inferior 
to hickory in rigidity by a factor of only 1.9. Pine differs 
from hickory not so much in stiffness as in brittleness — 
pine breaks where hickory bends. Spruce, of all the com- 
mon woods, has the highest strength for its weight. 

The tensile strength of wood varies in different directions. 
Along the grain its strength is ordinarily from ten to twenty 
times as great as it is across the grain. Also, the modulus of 



Properties of Wood 



Per- Sugar 




simmon Maple 



Specific gravity 






Static bending: fiber stress at 

elastic limit (1000 lbs. per sq. 








Static bending: modulus of rup- 

ture (1000 lbs. per sq. inch) . . . 






Static bending: modulus of elas- 

ticity (1,000,000 lbs. per sq. 







Static bending: work in bending 

to maximum load (lbs. per cu. 







Impact bending: energy .of 

dropped hammer to cause 

complete failure (relative) .... 






Compressional fiber stress at the 

elastic limit parallel to the 

grain (1000 lbs. per sq. inch) . . 





Maximum crushing strength 

(1000 lbs. per sq. inch) 






Compressional fiber stress at the 

elastic limit (perpendicular to 

the grain) (1000 lbs.' per sq. 







Tensional strength perpendicular 

to the grain (1000 lbs. per sq. 






Shearing strength parallel to the 

grain (1000 lbs. per sq. inch) . . 






Hardness on-the side (relative) . . 





elasticity is correspondingly greater along the grain. This 
anisotropy is avoided in plywoods, formed by gluing together 
three, five, seven, or nine layers of wood, the consecutive 
layers being arranged with their grain axes lying mutually 
perpendicular. Plywoods with the greatest number of 
layers are most resistant to splitting and are most nearly 
isotropic. The thick plywoods use cores of wormy chestnut. 
The ease with which woods are cut and carved is proportional 


to their homogeneity or the degree of similarity between the 
physical properties of the spring and the summer growth. 
Also, it is desirable to have a fine grain, a quality possessed 
by many hardwoods, especially mahogany. Of the common 
softwoods, poplar and sugar pine are the most homogeneous 
and easiest to work. 

Some wood substitutes are now available which are nearly 
isotropic. These are formed of bonded cellulose fibers. 
Although they are quite homogeneous, they are not so easily 
worked as wood with the plane and chisel, and nailing splits 
them. They can be sawed with the ordinary wood saw. 
Masonite is an example of such a wood substitute. 12 It 
comes, chiefly, in three grades, a light material which is a 
good heat insulator, a harder material which is suitable for 
making boxes for instruments, and an oil-tempered water- 
proof material. 

Waxes and cements. The physicist uses waxes and 
cements to seal windows into apparatus, tubes in plates, 
tubes together, and so forth. He uses them also to support 
and fasten down lenses, prisms, and mirrors. Of all waxes 
available, the most useful for making improvised supports 
and seals is the so-called universal wax. 

Universal wax. Universal wax is made from 1 part 
Venetian turpentine and 5 parts beeswax. It is usually, 
although not necessarily, colored with vermilion. It should 
be made up in small quantities, for it oxidizes, with the result 
that it becomes hard and loses its desirable properties. Old 
pieces may be useful if the outside oxidized layers are re- 
moved and discarded. The usefulness of this wax depends 
upon its adhesive and plastic properties. It is quite plastic 
at the slightly elevated temperature attained when the wax 
is worked between the fingers. When it cools, it becomes 
fairly rigid. 

Beeswax and rosin. Beeswax and rosin compound is pre- 
pared by melting together equal parts of beeswax and rosin. 
Its softening point is at the temperature which just begins to 

12 Masonite Corporation, 111 West Washington Street, Chicago, Illinois. 


feel hot (47°C.) and it is liquid at 10° above this temperature. 
Its outstanding property is its adhesiveness to cold metal. 
It is not very strong, but its strength is adequate for sealing 
vacuum systems and for fixing apparatus, as, for example, 
fastening a prism to the prism table of a spectrometer. It 
can be applied with a brush, an eye-dropper, or the blade of 
a knife. To secure the best bond to cold metal, the wax 
should be applied smoking hot with an eye-dropper or a 
knife. When it has been used for sealing down a bell jar, 
it can be removed with a putty knife, remelted, and used 
over and over. The smoking temperature distills off some 
of the beeswax, causing the compound to become harder. It 
may be re tempered by adding more beeswax. There are 
many applications for which this wax is not suitable because 
it shrinks a great deal on solidifying. It is best " dissolved" 
by a mixture of equal parts of carbon tetrachloride and ethyl 

Shellac. In its pure state, shellac in stick form is known as 
lapidarist's cement. It has a high tensile strength and shear 
strength. (Both are about 3800 lbs. /square inch.) Only 
the natural orange shellac possesses this high strength. The 
main ingredient of the better grades of sealing wax and 
especially banker's wax is shellac. 

Shellac is used in commerce chiefly for the manufacture of 
phonograph records, varnishes, and as an insulator in the 
electrical industry. It has a higher resilience than almost 
any other wax, and it is this property which gives long life to 
phonograph records. 

The best solvent for shellac is alcohol. This solution 
yields a varnish which has many uses in the laboratory. 
When it is very thick, it is useful for hunting leaks in vacuum 

Shellac is polymerized by heat, giving a product which is 
harder, has a higher softening temperature, and is less soluble 
in alcohol than the uncured material. This polymerization 
is accompanied by a chemical loss of water and a two- to 
threefold increase in molecular weight. Half of the uncured 


shellac is transformed into this harder variety by heating for 
30 hours at 90°C; at 150°C. it is completely transformed in 
3 hours. When the pure shellac is to be used as a cement, it 
is desirable to have it in the unpolymerized state. 

Commercial shellac may legally be designated as pure 
although it may contain as much as 3 per cent rosin. This 
materially weakens it. It is possible, however, to obtain 
shellac which is free from this adulterant. 13 

Tempered shellac. When shellac is tempered with 20 to 
40 per cent wood tar, we have a wax similar to the familiar 
DeKhotinsky cement. This wax is not affected by water, 
carbon disulphide, benzol, petroleum benzine, or turpentine. 
It is affected only slightly by ether, chloroform, and sul- 
phuric, nitric, or hydrochloric acids. 

When DeKhotinsky cement is heated in a flame, it emits 
an odor and is somewhat inflammable. A new variety of 
tempered shellac, which has no odor and is not so inflam- 
mable, is now sold by the Central Scientific Company under 
the trade name of Sealstix. Sealstix has a greater working 
range of temperature than pure shellac and a very high 

Shellac can be tempered with butyl phthalate. The re- 
sulting compound has a very low vapor pressure and is 
particuarly suitable for high-vacuum work. It is odorless 
and relatively noninflammable. 

Shellac can also be tempered to varying degrees with oil of 
cassia. About 10 per cent oil is quickly added to the molten 
shellac. The oil gives a compound with an agreeable odor. 
It is useful^f or many purposes when its vapor pressure is not 

Shellac can also be tempered with amyl acetate for use 
when the vapor pressure of this constituent is unobjection- 
able. Most of this solvent evaporates when the cemented 
elements are maintained at an elevated temperature (80°C.) 
for an hour or so. A mixture of 2 ounces of amyl acetate to 

13 Pure orange shellac is obtainable from William Zinsser and Company, 
516 West 49th Street, New York City. 


100 g of shellac gives a cement with a strength in excess of 
2500 lbs. /square inch. 14 

Picein. This sealing compound is characterized by low 
vapor pressure, plasticity at room temperature, and chemical 
inertness. Its low working temperature (it becomes quite 
plastic at 50°C. and is liquid at 80°C), together with its 
adhesiveness, recommends it for many applications. Be- 
sides its use for sealing tubes together and repairing leaks in 
vacuum systems, it is also used in the optical industry. It is 
practically unaffected by alcohol. Picein is immune even to 
a short immersion in cold dichromate cleaning solution. It 
is dissolved by benzol and turpentine. Its insulating quali- 
ties are said to be as good as amber if it is not overheated. 
It comes in two grades, the second being characterized by a 
liquefying temperature of 105°C. :I5 

Apiezon compounds. 16 Apiezon compounds are especially 
refined residues of paraffin oils freed from high vapor pressure 

The sealing compound "Q" contains graphite. It is 
plastic at ordinary temperatures and has a vapor pressure of 
10 -4 mm at room temperature, and, applied to ordinary 
twine, it is recommended as a packing for vacuum valves. 

Apiezon wax " W" has the lowest vapor pressure of any of 
the waxes now available. It is necessary to heat this wax to 
180°C. in order to raise its vapor pressure to 10 -3 mm of 
mercury. It melts at 70°C, but it can best be applied at 
100°C. or higher. Molten, it wets metals and glass and is 
quite fluid. It is fairly strong at ordinary temperatures. 
It is soluble in zylene. 

Silver chloride. Silver chloride is recommended for seals 
that must hold at elevated temperatures. It melts at 455°C. 
It is insoluble in water, alcohol, benzol, and acid. It is, 

14 This cement was developed by Marcus H. Brown. 

15 Picein and a rubber packing material, Dichtungsgummie, are obtainable 
from the distributing agents of the New York Hamburg Rubber Company, 
Schrader and Ehlers, 239 Fourth Avenue, New York City. 

16 Apiezon compounds are obtainable from the James G. Biddle Company 
Philadelphia, Pennsylvania. 


however, soluble in a solution of sodium thiosulphate. Most 
metals and glasses are wet by fused silver chloride. It is 
useful for sealing optically worked windows on a discharge 
tube. The window, after being sealed, is cooled slowly to 
prevent it from cracking. 

Espe and Knoll describe an enamel which the}' recommend 
for cementing optical plane parallels on a discharge tube. 17 
This is a mixture of clay and boracic acid, the melting point 
of which is 450° to 600°C. It is applied, as is silver chloride, 
by heating both the window and discharge tube in an 
electric oven. 

The bonding materials which we have considered above 
are thermoplastics. With the exception of shellac, the 
changes in their properties are reversible with temperature. 
We will now treat those substances which set, which can be 
vulcanized, and which polymerize by the application of heat. 
They include the synthetic resins, rubber cements, and 
inorganic cements. 

Synthetic resins treated of here fall into three broad 
divisions. These are, first, the polymerized phenol alde- 
hydes, of which Bakelite is an example; second, the con- 
densation products formed by polyhydric alcohols with poly- 
basic acids (these are termed alkyd resins, of which Glyptal 
is an example); and third, the polymerized derivatives of 
methacrylic acid, of which Lucite and Plexiglas are examples. 

Bakelite. 18 Bakelite comes in several forms that are useful 
to the physicist. The properties of these vary from liquid 
or soluble solids in the uncured condition to stable insoluble 
solids in the cured condition. Bakelite in the latter con- 
dition is obtainable in the form of clear, transparent sheets, 
blocks, tubes, and so forth. This material is light (density, 
1.27) and strong (7000 lbs. /square inch), is a good electrical 
insulator, and is insensitive to moderate heat. In this com- 
pletely polymerized form it does not melt, and it chars only 

17 Espe, W., and Knoll, M., Werkstoffkundeder Hochvakuumtechnik, page 157. 
Berlin: Julius Springer, 1936. 

18 Bakelite is manufactured by the Bakelite Corporation of America, 
247 Park Avenue, New York City. 


at a temperature of 285°C. Chemically, it is relatively inert. 
The completely polymerized Bakelite is unaffected by hot 
water, oils, greases, alcohol, acetone, benzene, dilute mineral 
acids, including hydrofluoric, and soap. It is practically 
nonhygroscopic. These properties recommend it as a 
material for making transparent chemical apparatus, such as 
burettes, pipettes, beakers, and so forth. Transparent 
forms of Bakelite are suitable for making models for photo- 
elastic studies with polarized light. . 

Several molded and laminated products bonded with 
Bakelite are available commercially. These have canvas, 
wood fiber, asbestos, or graphite as a base. The asbestos- 
base material is especially heat resistant, and the graphite- 
base material is useful for dry bearings. 

Bakelite varnishes usually consist of solutions of the 
unpolymerized form. After application and drying, the 
varnish films are transformed to the insoluble form by 

Bakelite cements come in the form of solids and viscous 
liquids. The solid form melts at about 80°C. (in hot water) 
and is transformed by heat to a form which does not melt. 
The liquid forms contain a volatile solvent. It is first 
necessary to evaporate this by preliminary heating of 1 to 
4 hours at 80°C, after which the residue is polymerized by 
heating for 2 hours at 120°C. A self -hardening cement is 
available which will set at room temperature. Vacuum seals 
made with these cements have a low vapor pressure and can 
be used to temperatures slightly above 100°C. 

A general-utility cement can be made by mixing Bakelite 
varnish with red lead. This hardens rapidly and will with- 
stand high pressure, steam, oil, and moderate heating. 

Alkyd resins. 19 Alkyd resins are formed by the conden- 
sation of phthalic anhydride on glycol, glycerol, or other 
polyhydric alcohols. Glycol phthalate is useful as a 
vacuum-sealing cement because of its low vapor pressure, 
fluidity, and wetting power when melted. In addition, it 

19 Alkyd resins may be obtained from the General Electric Company. 


may be cured to give it increased strength and inertness. It 
is also noted for its adhesiveness to aluminum. It is inert 
toward mineral oils. Dehydrating catalysts, such as zinc 
oxide, hasten the cure of these compounds and serve as a 
filler to economize on the resin, as, for example, in lamp- 
basing cements. 

Lucite and Plexiglas. Lucite and Plexiglas are trade 
names for methyl (and ethyl) methacrylate, polymerized 
derivatives of methacrylic acid. 20 These materials are sold 
as a cast resin in the form of sheets, rods, and tubes, as a 
thermoplastic powder, and as the unpolymerized liquid. 

The methyl methacrylate monomer is a mobile liquid 
which can be polymerized in almost any desired form. The 
monomer boils at 100°C. and has a heat of polymerization of 
about 80 calories /g. As it is obtained from the factory, it 
contains an inhibitor, such as hydroquinone or pyrogallol, to 
prevent it from polymerizing at room temperature. To use 
the liquid, this inhibitor is removed by washing with caustic, 
the liquid is dried, and an accelerator, usually benzoyl 
peroxide, is added to catalyze the polymerization. The 
volume of the monomer is 20 per cent greater than the 
volume of the polymer finally obtained, so that considerable 
art must be invoked to prevent the formation of voids when 
the monomer condenses. 

The polymerized methyl methacrylate, Lucite, has the 
optical properties given in Table VII. (See also Table VI, 
Chapter IX.) The polymers are inert toward water and in- 
soluble in the straight chain hydrocarbons, alcohols, and 
ethers and in most fats, oils, and waxes. They are, however, 
dissolved by lower ketone and ester solvents, and by mix- 
tures of the aromatic hydrocarbons when small amounts of 
alcohol are added. 

These polymers are vastly different from Bakelite and the 
plastic, Catalin, in respect to cutting. Whereas Catalin and 

20 Lucite is manufactured by E. I. duPont de Nemours and Company, 
Wilmington, Delaware. Plexiglas is manufactured by Rohm and Haas 
Company, 222 West Washington Square, Philadelphia, Pennsylvania. 




Percentage of Ultraviolet Transmission of Lucite at Different 

Wave Lengths 

Wave Length 



Sheet Thickness 1 inch 

Sheet Thickness 0.1 inch 


















Bakelite quickly remove the edge from high-speed steel (in 
fact, cold-rolled steel is just about as good for cutting them 
as high-speed steel), Lucite and Plexiglas can be cut by the 
hour without the edge of the tool becoming dulled. 

Fish-glue cement. A cement which is inert toward most 
organic solvents is made from a thick solution of 3 ounces of 
fish glue, i ounce of potassium bichromate, and a little am- 
monia. The cement so formed is allowed to dry and is then 
heated in an air oven until it assumes a chocolate-brown 
color. This cement is often used on Pulfrich refractometers. 

Rubber cements. Rubber cements are conveniently classi- 
fied as follows: non vulcanizing cements, which attain their 
strength simply by the evaporation of a solvent; vulcanizing 
cements, in which a chemical change occurs after the evapo- 
ration of the solvent; and thermoplastic cements. Some of 
the vulcanizing cements contain sulphur, while others are 
vulcanized simply by painting a vulcanizing liquid, sulphur 
chloride, on the rubber after it has been applied. 

The synthetic thermoplastic rubber-like products Neo- 
prene (manufactured by DuPont Company) and Koroseal 
(manufactured by the Goodrich Rubber Company) have 
many useful properties. These materials are remarkably 


stable chemically; they are inert toward acids and alkalies, 
as well as many fats and oils. 

Plaster of Paris. This is frequently used to support large 
glass bulbs containing mercury. The plaster suspended in 
water to the consistency of a paste is cast between the bulb 
and a loose-fitting wooden support. Salt shortens the time 
required for plaster of Paris to set, while a trace of glue acts 
in the opposite way. The glass may first be wiped with oil 
so that the plaster will not adhere to it. This facilitates 
subsequent dismantling. 

Litharge and glycerin. This combination gives a cement 
useful for the same type of applications for which plaster of 
Paris is useful. It is inert toward water, most acids, and all 
alkalies, and holds up to temperatures of 260°C. It is 
prepared by mixing pulverized Litharge (which has been 
first thoroughly heated at 400°C.) with pure glycerin to the 
consistency of a paste. 21 

Other irreversible cements. Water glass forms cements 
when mixed with the carbonates or oxides of calcium, mag- 
nesium, zinc, lead, or iron. In a few hours these mixtures 
set to rock hardness. Combined with talc, water glass 
makes a cement which holds even at a red heat. This 
cement will not chip off from glass at liquid air temperatures. 

Zinc oxychloride cement is used extensively in dentistry. 
It is formed from a 60 per cent zinc chloride solution and 
zinc oxide powder mixed to the consistency of a thick paste. 
These constituents react to give zinc oxychloride. To 
insure that the oxide is free from carbonate, it should first be 
heated until it turns yellow to calcine the carbonates. 

Nine parts kaolin mixed with one part borax give a cement 
useful to 1600°C. The constituent powders are mixed, and 
water is added to facilitate application. After the water 
evaporates, the cement is slowly heated to a yellow heat in 
order to set it. 

Insa-lute cement, a commercial product, is a thick white 

21 von Angerer, Ernst, Technische Kunstgriffe hex physikalischen Unier- 
suchungen. Friedr. Vieweg und Sohn, 1936. 


suspension of refractory substance in water glass. It sets on 
drying to form a white material having the texture of porce- 
lain. It is an electrical insulator and stands firing to about 
1100°F. It adheres to metal, glass, and porcelain. It 
attacks chromium-alloy wire at elevated temperatures and 
should not be used in contact with it. One should use a 
refractory cement such as Alundum cement in contact with 
chromium-alloy wires. 

Glue. Unquestionably, the best bonding material for 
wood is glue. Glues are more effective for the lighter woods, 
which contain less oils and resins, than for the dense woods. 
There are three kinds of glue: casein, blood albumin, and 
animal glues. The first two are useful for general construc- 
tion work. The animal glues exhibit greater strength, but 
they are softened by moisture. Casein glue is made from 
milk protein and lime. Blood glues contain caustic soda and 
water glass. Both the latter glues require heat and pressure 
for their application. They are water resistant and are used 
for making plywood. 

Animal glue is best applied hot. Cabinet and pattern 
makers usually keep a hot glue pot. For occasional use, 
however, air-drying glues are quite satisfactory for joining 
wood as well as leather. Air-drying glue is applied to the 
surfaces which are to be fastened together. The glue films 
on these surfaces are allowed to dry until the glue is definitely 
stringy. At this stage the surfaces are clamped together, 
and the glue is allowed to become completely dry. 

Lubrication. There are two kinds of lubrication with 
liquids. In the first and most common kind, called com- 
plete lubrication, the bearing surfaces are separated by a 
layer of oil about .005 inch in thickness. The friction, and 
consequently the amount of heat produced in the lubricant, 
depend on the thickness and viscosity of the liquid. 

In the second kind of lubrication the surfaces are in con- 
tact. Friction and galling are diminished and prevented by 
an absorbed surface film. The tenacity with which this film 
adheres and the effectiveness with which it reduces friction 



are determined by a quality called the oiliness of the liquid. 
The lubrication of surfaces in contact is called boundary